DOC: review notebook weights in GLM

[josef-pkt]

notebook for weights in GLM seems to have an outdated section

https://www.statsmodels.org/dev/examples/notebooks/generated/glm_weights.html#aggregated-data:-exposure-and-var_weights

Note: LR test agrees with original observations, pearson_chi2 differs and has the wrong sign.
plus investigation afterwards

based on a quick look pearson_chi2 is correct and agrees across equivalent representation with or without weights.
(need to go through notebook to check)