Refs #9059

If an algorithm is never run it is a bit of a waste to have things created
and destroyed when registering them.
Refs #8784

Mainly removed Poco includes and used forward declarations instead.
Unsuprisingly this has affected some algorithms.
Refs #8784

to algorithm base classes.
Refs #4218

Also some const-correctness.

instead set it once from the config value when the framework starts. Also
exposed the setters to python to give greater control at runtime.

Note that there are some minor modifications to the accessibility of some of the member functions on the Algorithm class. This is to facilitate reuse with the MultiPeriodGroupAlgorithm sub-class, which requires similar functionality.

modified: Algorithm.h
modified: Algorithm.cpp

to WorkspaceHistory, for code reuse

Setting them as parents causes issues with the workspace locking
and really isn't the right thing to do.

API side and also refs #4787 added to LoadLiveData

Touches nearly everything unfortunately but is necessary.

The leak was due to the workspace history exponentially increasing
in size. This history was duplicated with each run and then one algorithm
added. The history is now in a set and only those algorithms that have not
been run before are added.

4 tests fail at this point

Also refs #996 cleanup

Also Refs #996: starting to clean up Algorithm.execute()

Refs #3620: Progress reporting can give fractions of a percent; also can show estimated time to algorithm completion. Refs #3557: Better progress reporting in MergeMDEW.

1) I don't think anything should be done here as fast algorithms may not be possible to cancel anyway.
2) Made Algorithm::cancel() method virtual and overridden it in Load to call cancel() on the concrete loader. A test file can be sans2d00005512.nxs.
3) MantidPlot makes sure that the AlgorithmDockWidget is visible when an algorithm is running. fixes #1733

Adds the option to select which algorithm categories are available into the Mantid->options tab of the preferences window.

Known Bug:  Does not work with nested categories at the moment, but right now the only one is DataHandling, and you'd be a fool to hide that.

The algorithm base class can now handle algorithms being in multiple categories, however this should not be used yet, and mantidplot or the python help system knows nothing about this.

Refs #3157 Sort our System.h. It was becoming a dumping ground. It should only be touched when absolutely necessary as it affects every file in Mantid. Also combined this with Refs #2345 to sort out DLLExport/DLLImport usage. It's done for Kernel, Geometry and API but stuff in the plugins could do with updating as well.

Refs #2897. Actually give toString a default so that it makes sense in most situations. This will fix the Python test failure.

Refs #2987. Fix issues with some code I wrote a while a go that shows up as a a warning on the Mac. Also cleaned up another prevalent warning from IMDGeometry.

Refs #3026:  Added a field for Wiki description to Algorithm. Made WikiMaker use it to generate wiki pages.