Refactor sample return

[michaelosthege]

This refactors the post-MCMC part of pm.sample into a function which simplifies testing and future refactoring.

In test_mcmc.py I consolidated the test cases, mostly by using pm.Metropolis and avoiding to re-run MCMCs just to test different flavors of discard_tuned_samples, return_inferencedata and so on.

Locally this reduced the test_mcmc.py runtime by ~200 seconds.

Maintenance

• Post-MCMC parts of pm.sample were extracted into a function.

[codecov]

Codecov Report

Merging #6546 (6a2ed47) into main (9b59771) will decrease coverage by 0.05%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6546      +/-   ##
==========================================
- Coverage   91.96%   91.91%   -0.05%     
==========================================
  Files          89       89              
  Lines       14935    14936       +1     
==========================================
- Hits        13735    13729       -6     
- Misses       1200     1207       +7     

Impacted Files	Coverage Δ
pymc/sampling/mcmc.py	92.12% <100.00%> (-0.22%)	⬇️
pymc/stats/convergence.py	90.35% <0.00%> (-5.27%)	⬇️

[Armavica]

Nice!