Issue warning if RVs are present in derived probability graphs

[ricardoV94]

This is problematic when using Potentials with derived quantities expressions, such as in:

with pm.Model() as m:
  x = pm.Normal("x")
  y = pm.Potential("y", pm.logp(pm.Normal.dist(x).exp(), 1.0))  # raises new warning
  m.logp()  # contains a random variable (clone of x)

Where x is not be replaced by the value variable in the Potential expression of y, leading to the wrong model logp.

I couldn't think of a clean solution, so we just give an informative warning about the problem and suggestion on how to fix it.

Note that this is not a problem in CustomDist because the logp receives the value variables already instead of the random inputs.

[codecov]

Codecov Report

Merging #6651 (f9e48e0) into main (b7764dd) will decrease coverage by 0.49%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6651      +/-   ##
==========================================
- Coverage   91.96%   91.48%   -0.49%     
==========================================
  Files          94       94              
  Lines       15927    15942      +15     
==========================================
- Hits        14647    14584      -63     
- Misses       1280     1358      +78     

Impacted Files	Coverage Δ
pymc/logprob/basic.py	99.15% <100.00%> (+0.12%)	⬆️

... and 6 files with indirect coverage changes

[michaelosthege]

LGTM