Allow algorithms to be "ReductionSteps" - name to be improved

[Nick Draper]

This requires algorithms to not store themselves in the history of a workspace.

It was also discussed that they should store a list of the algorithms to be used and the options of underlying algorithms.

[Mathieu Doucet]

(In [8858]) Modified python algorithm functions to allow algorithm object to be created without being run. Re #2129

[Mathieu Doucet]

From python, we can now create an algorithm object without executing it by calling

 algo = [algorithm name]( [parameters...], execute=False)

For example:

 algo = Load("myfile.txt", "mywksp", execute=False)

Once created, the algorithm can be run by doing

mantidsimple.execute_algorithm(algo)

The Reducer has been modified so that Mantid algorithms can be passed to append_step(). Previously, this method only accepted ReductionStep objects.

r = Reducer()
r.append_step(Load, "myfile.txt", "mywksp")

The Reducer will create the algorithm and only execute it when reduce() is called. Note that ReductionStep objects can still be passed to append_step() and this change is fully backward compatible. Also note that the input and output workspaces, when defined, will be over-written by the Reducer. Still, dummy values should be passed according to the algorithm's syntax.

[Mathieu Doucet]

(In [8865]) Minor tweak to mantidsimple to allow not to set mandatory properties when the algo is not to be executed immediately. Re #2129

[Mathieu Doucet]

Any Reducer method with a signature of the type

 def some_method(reduction_step)

where reduction_step is a ReductionStep object, can now take a mantidsimple algorithm function and its parameters by decorating the method with @validate_step

 @validate_step
 def some_method(reduction_step)

[Mathieu Doucet]

(In [8866]) Make sure a new algorithm is created every time a reduction step is executed. Re #2129

[Mathieu Doucet]

(In [8867]) Cleaning up. Re #2129

[Martyn Gigg]

(In [8872]) Minor fix for reducer module so that it is compatible with python 2.5. Apparently 2.5 doesn't accept something sensible like func(*args, execute=False, kwargs) while 2.6 is perfectly happy with this. Refs #2129

[Mathieu Doucet]

(In [8890]) validate_step decorator will now work for any combination of mandatory/optional arguments. Re #2129

[Mathieu Doucet]

Needs unit tests, otherwise the needed functionality should be complete. I have created a new ticket (#2270) to deal with history logging issue of python sub-algorithms.

[Mathieu Doucet]

(In [8907]) Added unit tests. Re #2129

[Mathieu Doucet]

(In [8910]) Forgot mtd.initialise(). Re #2129

[Mathieu Doucet]

(In [8912]) Added test. Re #2129

[Michael Whitty]

tested in [9528] - allows base algorithms to be set as reduction steps.

[Stuart Campbell]

This ticket has been transferred to github issue ​2976