[Proposal] Support for glmtoolbox glmgee objects (glmgee or gnmgee models) / interim warnings

Performance does a fantastic job of bringing model checks under one hood. I'd be very pleased to see support expanded towards glmgee objects from the glmtoolbox package. The benefits of this GEE implementation have been highlighted by @adrianolszewski in discussions over at marginaleffects/issues/1148, and some performance functions already work with it (for example, check_collinearity). However, other aspects are either unsupported, or run but may not actually be doing what the user might expect. For example, check_outliers runs, but it's not clear whether this is taking into account the cluster correlation aspects of the model. There are some useful residuals and influence methods already implemented in glmtoolbox, residuals.glmgee with Mahalanobis and Pearson options, and cooks.distance.glmgee, and it would be great to have specific support within performance. An interim warning in the same manner as #95 could be applied in the meantime, if this functionality is not yet included.

Thanks again for this awesome tool!

Start by reviewing existing glmgee handling in check_collinearity and check_outliers, then inspect glmtoolbox's residuals.glmgee and cooks.distance.glmgee methods. Compare the current behavior with the cluster-correlation requirements described in this issue. Done means either adding reliable glmgee diagnostics or documenting the unsupported cases and adding an interim warning like #95.