The minimum is set to Python 3.6 so this can be assumed

Avoids cmake warning that looks like an error.

Upstream libraries no longer generate warnings so cppcheck
generates an unmatched suppression warning. Remove this.

On Windows, even though certain things are cached, running
these finders multiple times is costly. Put everything
at the top level. Our current CMake structure does not really
offer options fine-grained control of enabled components
so there is no benefit to the finders being in multiple
places.

It has not been used in years and our licence has expired.

It can find all components in one shot and offers
import targets over the traditional global variable
approach.

Adds boost-python to public interface of PythonInterfaceCore
Removes the need for each library to link to it by hand and
it only needs to be found once.

As we now require C++17 we must have at least gcc 7

This no longer works and has been unused for some time.
It is superceded by the ENABLE_MANTIDPLOT & ENABLE_WORKBENCH
flags.

Names to find_package_handle_standard_args now need to
match the call to find_package exactly

If qwt 6 was installed the finder would find the headers
for Qwt5 but library for Qwt6. This forces the library
finder to only find version 5 of the library.

Also removes MANTIDLIBS as a search location as these are
now ancient history.

CMake expected the package to be called Qwt5 but this
did not match that given to find_package_handle_standard_args

FindPython3 now comes with the ability to find Numpy
so that we can drop some of our finder code

`LRAutoReduction` can now use `LRReductionWithReference`

Fixes for sliceviewer line plots

Fix overplotting resetting axes scales

Fix workbench tests for Qt 5.14

Fix bug in MonteCarloAbsorption where spectra have diff wavelength ranges

Co-authored-by: PhilColebrooke <52415735+PhilColebrooke@users.noreply.github.com>

A widget cannot be managed by 2 separate layouts.

Sliceviewer tests now all run in under a second

Zoom is actually just a different set of axes limits
from that specified by the image itself. Keep these
limits, taking into account transposition, when
swapping dimensions.

Reset to "home now queries dimensions for the real
extent of the data to set the appropriate axes limits

If the dimensions are flipped while the view is zoomed
or panned then it is unclear what area a user would
wish to see and its possible the new area is outside
of the image. For this reason reset the view limits
to the data limits on making a new plot.

A zoom/pan usually changes both the X/Y limits and
resulted in the matrix image being resampled twice
for those events. This should now be lazily delayed
until just before drawing.
Refs #28579

The lazy instantiaion of the peaks presenter meant
that the first call to the property created the
object even in instances where the peaks button
had not been pressed or it was disabled!
This manifested itself on the view for matrix workspaces
constantly responding to resize events that inserting
an empty peaks view would create.
Refs #28579

Refs #28579

This avoids a frozen line plot where it is unclear what point
these lines would correspond to. It then avoids an error
when moving the mouse to flip the axes

The removed image keeps trying to update itself
if the events are not manually disconnected.

Set only image tick labels invisible when lineplots active
as setting the whole image X/Y axis invisible disables grid
functionality. We actually only need to make sure the
duplicate labels don't overlap the shared axis.