WIP: persist pickles for TOAs loaded during a test run

[aarchiba]

Currently many of our tests load TOAs from files in tests/datafile/. Some of these files are quite large. PINT has a facility for making re-loading of tim files much faster: saving a pickle in which clock corrections and similar data have already been computed. This PR is intended to make the test suite use this facility, where appropriate, by creating these pickles in a temporary directory that persists for the duration of a test run.

This PR also add the ability to easily specify a directory in which pickles should be kept, named according to the name of the .tim files they are made from. This may be useful in other contexts.

This PR needs to convert many old unittest-style test cases so that they can use pytest fixtures. Minor test cleanups may also happen.

[aarchiba]

This PR is inevitably going to conflict with #1305 so perhaps best not to do too much until that is merged.

[codecov]

Codecov Report

Merging #1336 (5553eba) into master (24e144a) will decrease coverage by 0.04%.
The diff coverage is 91.66%.

@@            Coverage Diff             @@
##           master    #1336      +/-   ##
==========================================
- Coverage   61.31%   61.27%   -0.05%     
==========================================
  Files          89       90       +1     
  Lines       20218    20366     +148     
  Branches     3619     3653      +34     
==========================================
+ Hits        12397    12479      +82     
- Misses       7046     7103      +57     
- Partials      775      784       +9     

Impacted Files	Coverage Δ
src/pint/polycos.py	77.19% <ø> (+0.32%)	⬆️
src/pint/models/binary_dd.py	75.86% <33.33%> (ø)
src/pint/models/noise_model.py	86.85% <100.00%> (+0.10%)	⬆️
src/pint/pint_matrix.py	67.62% <100.00%> (-0.53%)	⬇️
src/pint/toa.py	83.25% <100.00%> (+0.25%)	⬆️
src/pint/models/timing_model.py	81.63% <0.00%> (-1.77%)	⬇️
src/pint/models/parameter.py	81.48% <0.00%> (-0.25%)	⬇️
src/pint/scripts/compare_parfiles.py	0.00% <0.00%> (ø)
... and 1 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 24e144a...5553eba. Read the comment docs.

[aarchiba]

Key TODO item: compare full-test-suite runtime with pickle_dir working properly versus returning None. Should be faster, or at least not slower, but best to check.