Ensure invert_rotation_axis=True is only applied once to each goniometer model

[dagewa]

More generally, ensure manual geometry changes are only applied once to each unique model.

[codecov]

Codecov Report

Merging #2469 (34ead2b) into main (bcd31c6) will decrease coverage by 0.02%.
Report is 1 commits behind head on main.
The diff coverage is 89.87%.

❗ Current head 34ead2b differs from pull request most recent head 77516da. Consider uploading reports for the commit 77516da to get more accurate results

@@            Coverage Diff             @@
##             main    #2469      +/-   ##
==========================================
- Coverage   78.61%   78.60%   -0.02%     
==========================================
  Files         607      607              
  Lines       74339    74410      +71     
  Branches    10118    10130      +12     
==========================================
+ Hits        58445    58488      +43     
- Misses      13735    13759      +24     
- Partials     2159     2163       +4     

[dagewa]

@graeme-winter I realise I'm tagging you a lot for PR reviews lately, but you came up in the suggestions 😉

[graeme-winter]

On the balance I am fairly sure that the change set meets the brief of fixing the issue as the added test exercises the right code paths. Some concerns:

• the motif of if scan and scan not in thing scan = worries me slightly as I am nervous that it does not do what we may intend
• I am left with the idea that invert_rotation_axis is fundamentally wrong, because the other overrides are idempotent but this one is a toggle (which gives rise to the fundamental issue)

On the latter - I note we do not offer invert_beam or invert_detector_origin etc. - however I also recognise that the inverted rotation axis is by far and away the most common issue. Just left with a slight sense of unease, which maybe I should document in a subsequent issue.

The change set I believe meets the brief and so is a good addition to the code base, just raises questions for me.