Re #21742

Re #20399

The new calculation for the detector pixel size is more accurate, so the test results are slightly different.

Re #20399

Re #20120

- Fix an inconsistency when summing in Q where normalisation by
transmission run was not being done if the input workspace is already in
wavelength, whereas this normalisation is always done when summing in
lambda. Both methods now always normalise by transmission run.
- Fix an inconsistency when summing in Q where cropping of the output
workspace was not being done.

- Add unit tests for summing in Q with normalisation by:
  - direct beam
  - monitor
  - transmission run
  - correction algorithm
- Add a unit test to check IvsQ when summing in Q.

- Previously the unit tests were using 3 detectors at the same position.
This commit changes them to be at different positions, which is required
for meaningful tests for the summation in Q functionality.
- Some test results have therefore changed as a result of the new
detector positions. To limit the changes, I have used processing
instructions of "2" where only a single detector is used. This is the
centre detector and has the same position as it did previously.
- The order of the processing instructions in the tests for multiple detector groups have
been modified to check that the resulting groups are being sorted correctly.
- Unit tests for summing in Q added.

This adds a unit test for a case where spectrum numbers are different in
the transmission and run workspaces, but mapping between them is not
done because StrictSpectrumChecking is set to false.

This tests an attempt to normalise by a transmission run with strict
spectrum checking enabled but where the requested spectrum does not
exist in the transmission run

In this test, the transmission run has different spectrum indices to the
run. The spectrum numbers from the run are mapped to those in the
transmission run.

Note that some tests have been removed in ReflectometryReductionOne2Test. The reason is that those tests were testing the rebinned workspace in Q which no longer happens in this algorithm but in ReflectometryReductionOneAuto. Note that three tests were removed from ReflectometryReductionOneAuto2Test. The reason is that the algorithms wouldn't run anymore due to Rebin, which now has strict tests for rebin params. Rebin was failing in the removed tests because the input workspaces used were not appropriate for the reduction, for instance, the OFFSPEC was a scan run.