Moving data sorting from plotting to data input

[wathen]

Description of Work

Fixes part of #242

Moves data sorting out of fitbenchmarking/fitting/plotting to fitbenchmarking/parsing so that the data is sorted once.

Testing Instructions

1. Check example results from master

[tyronerees]

Checked on simple_tests and CUTEst (all minimizers) and Neutron (scipy and mantid), and no changes in the accuracy.

The times are around 50% quicker with sorted data, suggesting that the fitting routines sort the data first.

[tyronerees]

For some reason this change slowed things down quite a bit on my system. I didn’t have time to investigate, and won’t be able to follow until I’m back in a week. I’ll pass over reviewing duty to @AndrewLister-STFC so this doesn’t get held up

[AndrewLister-STFC]

I think indexing data_e might be dangerous as it could be None?

I'm going to check the timings that @tyronerees mentioned now.

[AndrewLister-STFC]

For some reason this change slowed things down quite a bit on my system. I didn’t have time to investigate, and won’t be able to follow until I’m back in a week. I’ll pass over reviewing duty to @AndrewLister-STFC so this doesn’t get held up

I had a look on mine, and found that the branch gives timings which are similar but slower.
As an example, I ran scipy against SasModelling/1D, simple_tests, and Muon. On the branch I get ~1min 10s and on master it's ~1min 3s