evaluate_transformer_performance crashes if any profiling method crashes

[amontanez24]

Problem Description

Currently the evaluate_transformer_performance method will profile the transformer provided against the dataset generator provided for three different sizes. It does this by running profile_transformer which runs fit, transform and reverse_transform for the transformer and times these methods as well as captures their peak memory. The problem is that if any of these individual profiling methods fails, the whole evaluate_transformer_performance method errors out. This means that if the profiling was able to complete for some of it, it ends up being lost.

Expected behavior

We should catch any errors thrown from profiling the memory or time for any method and just replace that with np.nan, so that the rest of the performance values can still be collected.