Implement compute_log_prior utility

[aloctavodia]

Add compute_log_prior function. the log_prior values are needed for prior sensitivity checks as described in this paper https://arxiv.org/abs/2107.14054. Part of the functionality described in that paper is already implemented in ArviZ.

---

📚 Documentation preview 📚: https://pymc--7149.org.readthedocs.build/en/7149/

[ricardoV94]

Can't this all be implemented in the same file?

[aloctavodia]

sure, we just need to rename log_likelihood.py to log_density.py or something similar and fix a few imports.

[codecov]

Codecov Report

Attention: 3 lines in your changes are missing coverage. Please review.

Comparison is base (627a8dd) 90.41% compared to head (47025f0) 91.92%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #7149      +/-   ##
==========================================
+ Coverage   90.41%   91.92%   +1.50%     
==========================================
  Files         101      101              
  Lines       16930    16945      +15     
==========================================
+ Hits        15308    15576     +268     
+ Misses       1622     1369     -253     

Files	Coverage Δ
pymc/stats/__init__.py	100.00% <100.00%> (ø)
pymc/stats/log_density.py	98.30% <96.55%> (ø)
pymc/backends/arviz.py	96.07% <60.00%> (-0.93%)	⬇️

... and 10 files with indirect coverage changes