Add _as_tensor_variable converter for pandas objects

[juanitorduz]

This is an attempt to solve #5877

@ricardoV94 I went through pymc/aesaraf.py but I could not figure out by myself where to add your suggestion. I decided to open a PR anyway but I would appreciate some hints on which function this has to be included and tested 🙏

[codecov]

Codecov Report

Merging #5920 (275ee6f) into main (0b4f248) will increase coverage by 0.02%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #5920      +/-   ##
==========================================
+ Coverage   89.53%   89.55%   +0.02%     
==========================================
  Files          73       73              
  Lines       13267    13271       +4     
==========================================
+ Hits        11878    11885       +7     
+ Misses       1389     1386       -3     

Impacted Files	Coverage Δ
pymc/aesaraf.py	92.19% <100.00%> (+0.11%)	⬆️
pymc/tuning/starting.py	92.43% <0.00%> (-0.13%)	⬇️
pymc/distributions/continuous.py	97.98% <0.00%> (ø)
pymc/distributions/timeseries.py	78.64% <0.00%> (ø)
pymc/parallel_sampling.py	86.79% <0.00%> (+0.99%)	⬆️

[michaelosthege]

ArviZ has a dependency on pandas already, so you can import it right away:

https://github.com/arviz-devs/arviz/blob/main/requirements.txt#L6

I'd still add it to the requirements and conda envs though.

A good placement for this function might be close to the top of the file, because there's already a function for converting observed data, which is slightly similar.

Any idea what to do with multi-indexed DataFrames?

[ricardoV94]

Oh, we removed pandas dependency on purpose but we didn't check if any of our dependencies required it... That's a shame :/

[juanitorduz]

I think this first iteration is ready for review :)

[michaelosthege]

Looks good to me, but I'll ask in the chat for a second person to look at it & merge (assuming the pipeline goes green).

[michaelosthege]

The changes requested by @ricardoV94 were made.
I'll merge now.

Thanks for this contribution @juanitorduz !