Improve Laplace computation and integration

[aloctavodia]

• Perform computations on the unbounded space and return results on the original scale.
• Return InferenceData object

[codecov-commenter]

Codecov Report

Merging #555 (c1baf2f) into main (d84080b) will increase coverage by 0.29%.
The diff coverage is 97.43%.

@@            Coverage Diff             @@
##             main     #555      +/-   ##
==========================================
+ Coverage   85.72%   86.01%   +0.29%     
==========================================
  Files          32       32              
  Lines        2641     2689      +48     
==========================================
+ Hits         2264     2313      +49     
+ Misses        377      376       -1     

Impacted Files	Coverage Δ
bambi/models.py	85.25% <ø> (+0.18%)	⬆️
bambi/backend/pymc.py	82.04% <97.29%> (+1.00%)	⬆️
bambi/tests/test_built_models.py	96.14% <100.00%> (+0.46%)	⬆️
bambi/tests/test_data.py	100.00% <0.00%> (ø)
bambi/tests/test_priors.py	100.00% <0.00%> (ø)
bambi/tests/test_predict.py	100.00% <0.00%> (ø)
bambi/tests/test_model_construction.py	98.83% <0.00%> (+0.02%)	⬆️
bambi/backend/terms.py	96.25% <0.00%> (+0.09%)	⬆️
bambi/families/likelihood.py	93.61% <0.00%> (+0.28%)	⬆️
... and 2 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[tomicapretto]

This also makes the code in _run_laplace() much clearer, that's great.

One extra thing we could have is a docstring in _posterior_to_idata(). Maybe it also needs a better name, like _laplace_posterior_to_idata(), _posterior_samples_to_idata(), or something that is more explicit about the circumstances it's used. The previous name is too general.

[aloctavodia]

This is ready to merge from my side