Similarity Heatmap and Scatter Plot should warn users when running on large datasets

[JacksonBurns]

For datasets with more than a couple hundred molecules, the similarity heatmap is not helpful to look at and the embeddings needed for the scatter plot take far too long to generate. Automatically disable generation of these for large datsets.

[JacksonBurns]

Rather than disabling (which would cause inconsistent behavior based on dataset size, which sounds frustrating) we are going to send a warning to the user when large (500+) datasets are attempted.