Sometimes experimental data is provided in meV. There are also a
couple of "gotchas" for end-users in converting results with Mantid's
ConvertUnits dialogue: EMode must be set to Indirect, incident energy
is not relevant, and "DeltaE" is the choice for meV while
"DeltaE_inWavenumber" is the choice for cm-1.

As a convenience, provide an Abins option in familiar units. The
selected units will also be assumed for any experimental data file
provided to the same Abins call.

Some of the supporting infrastructure is left behind, but generally
the purpose here is to strip back the changes a bit. This branch still
covers iteration over angles, Instrument settings, Lagrange and
pushing a bunch of the Abins algorithm into a parent class. Quite a
lot as-is, would be good to review/merge while solving problems with 2D.

Work in progress, developing angle-resolved outputs. This version
succesfully produces .nxspe files, but

- writes a useless dummy workspace
- doesn't let user choose output file location
- MSLICE doesn't look great with these files due to hard-coded
  detector widths

Currently, Abins calculations for direct instruments will loop over
several incident energies and sum the results together.

It is not clear that this is especially useful, and has not been
directly requested by users. In an experimental scenario we would
expect several spectra to be collected at different incident energies,
and compared with simulations at the same energies.

This removes one level of looping/nesting from Abins, and may allow us
to further remove some of the redundant code separating 1-D and 2-D
runs.

- PyChop is used to compute resolution functions.
  - Rather than implement each instrument separately, a common
    PyChopInstrument class can be instantiated to instances that
    represent different instruments. The abstraction in
    get_instrument() has paid off!
- In addition, the instrument data in Pychop provides detector angle
  ranges
  - We sub-sample these ranges to give a reasonable amount of coverage.
  - This is not quite the same as what we would get in MSLICE, as we
    are able to place each sub-angle along its exact trajectory (no
    ambiguity within detectors).
  - Without sub-sampling, the results appear too sparse and are
    difficult to interpret.

- refactor management of angles, so that instruments are responsible
  for reporting the series of sampled angles. This allows a wider
  range of approaches as appropriate.

Pass the selected setting to get_instrument. For now this is only
implemented in Abins1D (for use with Lagrange).

The property setup is implemented in abinsalgorithm to enable a more
consistent infrastructure; we do expect 2D instruments to have settings.

- Add a Lagrange instrument, based on TOSCA
- For now, factor common features into a common parent class for
  indirect instruments supporting powder averaging. As model
  is developed it should become clear how similar they really are...
- At the moment the energy-q relationship is copied from TOSCA. In
  fact Lagrange has a different geometry that seems to sample many q
  for each final energy. More development/discussion needed.
- Add a "setting" option to Abins interface, selecting from instrument
  configurations. For Lagrange this selects the monochromator.
- The broadening is selected depending on the monochromator. For now
  the values follow the relations reported on Lagrange
  website. However, it should be possible to obtain a more
  accurate (and complex) energy/width function for use here.

- The bin_width advanced parameter, while no longer required for
  general program flow, is still needed for cache invalidation of HDF5
  data from previous runs with different sampling. This is restored to
  prevent Abins breaking when bin_width was changed in a follow-up
  run.

- Allow the Map2D instrument e_init to be set from the GUI. This is an
  adjustable parameter for many direct instruments, and should be
  easily accessible in user simulations. It strongly interacts with
  the sample angle to determine the measured spectrum.

- Output at NOTICE level was excessive, especially for Abins2D
- The Abins S calculator can now accept a mantid Process object to
  handle progress bar as appropriate.
- We don't know at beginning of Algorithm how many steps S calculator
  will need, but as this is potentially the most expensive step we
  reserve 70% of the progress bar space. This is then subdivided as
  appropriate to the calculation.
- For now we increment at each atom, data q-point and incident
  energy. Finer increments would be possible by splitting over
  order (or even as chunks of bands are processed), but this already
  provides quite satisfactory progress and should deliver meaningful
  time estimates.

The _atoms_data is only used in a few places; pass it as an argument
rather than attaching it to the Algorithm as a private property.

This is an initial implementation of a separate "Abins2D" interface,
based on the version built for a demonstration to ISIS Molecular
Spectroscopy scientists.

- Further user testing is needed to determine the appropriate feature
  set. Suggestions already under consideration include:

  - Instrument setting/chopper options
  - Incident energy selection

The free entry of resolution value should be removed and replaced with
a reasonable selection of instruments/setups with appropriate set
values. A natural first target will be MAPS.

Initial testing shows a curious behaviour in the current
implementation, which reduces overall intensity as the resolution
width is increased.

Use multiple inheritance(!) to set up some private methods that
provide the common stuff and keep 1D and 2D algorithms fairly lean
and easy to a) compare b) maintain

- Also removing check and corresponding test that resolution width must be
  less than 1. Why not? If the user wants a broad spectrum, let them
  have one. It's not for us to say that a large delta function is
  unreasonable.

- delta_width doesn't do anything, use the 'resolution' parameter for
  broadening width

- remove all references to pkt_per_peak, this is a relic of the old
  broadening setup

- For now, use 'interpolate' scheme as default for 2D
  map. As we are using a fixed sigma, this should be equivalent to a
  simple convolution; we can improve performance from here by adding a
  'convolution' scheme that only accepts scalar sigma.

- remove offset of first sample by bin_width. This must be a remnant
  of when frequency samples were located at the right side of the bin.

- spelling correction / consistency: rebined -> rebinned

- PEP8 spacing cleanup

A few things broke while rebasing the direct map prototype onto some
other refactoring.

Some unit tests are still broken due to additional advanced
parameters. It would be nice if only the _relevant_ advanced parameter
blocks were written out and tested against; TwoDMap parameters have no
bearing on TOSCA calculations.

The handling of bin_width is a bit messy. Need to decide if this is
and advanced parameter (living with the wavenumber range parameters in
AbinsParameters.sampling) or something that should be in the user
interface (and passed around as a variable). Advanced parameter seems
more consistent and sensible all-round (insofar as using advanced
parameters is sensible at all.)

Abins.py is getting the active instrument name by querying the
Instrument object, but this information may also be available as
self._instrument_name? Potential for more cleanup there.

Known regression: summing over 2D spectra doesn't work

Currently the outputs display quite nicely in the MantidPlot spectral
viewer but don't seem to work at all with workbench slice viewer.

Performance is also slow relative to the 3.9.2 code, in part due to
the new instability detection/threshold logic which spends a lot of
iterations finding an appropriate threshold.

Functionality to generate 2-d maps with Abins by selecting an
alternative "Instrument", building on work by Krzysztof Dymkowski

This is essentially a 3-way merge between the removed code from Mantid
3.9.2, a more recent prototype set of AbinsModules, and the current
Mantid code base. Some new development was needed to get Abins.py
working but otherwise this is left as intact as possible for testing,
benchmarking and further development.

Currently not to be used for production calculations. Succesfully
generates workspaces that can be viewed with the
MantidPlot spectrum viewer, but results are unverified.
Default precision of angular sweep has been reduced for speed.