swloop mex (part II)

[mducle]

This is the second part of the PR to refactor the core computation loop in spinwave into a C++ mex file.

It also modifies spinwave.m to have two new options:

• neutron_output - if true this will give Sperp in the output spectrum struct and will not save Sab for a small speed-up and large memory saving.
• fastmode - this implements the code that was in spinwavefast() - it turns on neutron_output and moreover ignores all negative energy modes thus halving the cross-section calculations and saving a bit more memory.

spinwavefast has also been removed and some addition unit tests added to cover the two new options. Both options are fully supported by the swloop mex file.

NOTE: This branch should be refactored when the other PR is merged.

[github-actions[bot]]

Test Results

    4 files  ± 0    106 suites  ±0   14m 24s :stopwatch: +23s   703 tests +19    685 :white_check_mark: +32  18 :zzz: ±0  0 :x:  - 1  1 946 runs  +40  1 910 :white_check_mark: +56  36 :zzz: ±0  0 :x:  - 4 

Results for commit bb79de05. ± Comparison against base commit 0c2f2944.

:recycle: This comment has been updated with latest results.

[mducle]

@RichardWaiteSTFC

I feel like anywhere that the new options neutron_output and fastmode should be exposed to any function calling spinwave with fit_mode - I'm thinking specifically of fitspec - should we do that in a separate PR (will also require extra validation on Evect being positive if fastmode etc.)

Yes, I think it's best to have this as another PR. I'll make an issue for it.