GAM compute_p_values option issue

[katiewerner]

I'm new to using this package so apologies in advance if this ends up being user error...

H2O version, Operating System and Environment
python 3.11, h2o 3.46.0.6, windows

Actual behavior
GAM model trained as H2OGeneralizedAdditiveEstimator() with arguments:

• solver='IRLSM' (per the docs)
• remove_collinear_columns=True (per the docs)
• compute_p_values=True (per the docs)
• and with no regularization (per R documentation I found https://rdrr.io/cran/h2o/man/h2o.gam.html)
  gives a ValueError if you call model.coef_with_p_values().

If you edit line 882 in model_base.py to allow 'glm' and 'gam' then that method call doesn't error out, but it still doesn't provide the actual p_values as desired/expected...

The docs indicate this method should be available for both GLM and GAM:
https://docs.h2o.ai/h2o/latest-stable/h2o-docs/data-science/algo-params/compute_p_values.html

Expected behavior
Similar to a GLM model fit in H2O, we'd expect the p-values to be calculated and displayed if model.coeff_p_values() is called.