This test was broken when the build servers moved from GSL v1 to v2.

This test has a different tolerance for the fit depending on the GSL
library used. As OSX is now on v2, the test needed to be changed to
match the corresponding library tolerance.

Allows user to select plotly or matplotlib

This is used for the Parallax algorithm which uses muParser.
muParser has changed the meaning of log in version 2.2.5 from log10 to ln (Thanks for finding this out, MartynGigg).
The build servers do not provide the same version of the muParser.
To prevent future confusion, we like to use ln instead of log.

The existing iq D22 reference workspace is valid and should be used

Refs #24364

Refs #24364

The X axis in the reference files now correspond to the vertical axis in the initial workspace.

Re #24355

Refs #24364

Refs #24364

- Add function descriptions
- Avoid using inbuilt type name 'list' as a variable name
- Fix typo where 'slice' was used as a variable name
- Use camalCase for function names
- Include spaces after commas for parameter lists
- Minor refactoring of quickRef function to reduce duplication
- Minor change to move some code out of twoAngleFit

Different choices of types for data. Nexus format only requires NXnumber
for a larger number of its fields.

Test V20 file added. This is the 10th iteraton of the input file

There seems to be instability across platforms in the output of fitting. To work around this for now, remove the offending workspaces from the reference and do an internal check of values to a more generous tolerance.

I've left the script largely as provided by the scientists but done some tidying to improve readability

The results are slightly different (e.g. 45222 IvsQ ~0.001) from the original workspaces due to IDF changes since they were originally created.

This is intended to fix the failing conda systemtest build which is
framework only. The other changes are to reduce pep8/flake8 warnings.