Warn users when sample size is low (compared to number of features)

[sepro]

When the number of available samples is relatively low, while a lot of features are included the glm might throw warnings, you can get strange output (negative R2 values), ...

Catching these and providing info on what to do could be useful

[sepro]

Issue is that "low sample sizes" is depending on the data presented .... Haven't found a good heuristic to catch this (apart form maybe a warning when effect sizes drop below zero which could also just be a feature with little variability)