Documentation: highlight that from_dataframe is not for large datasets

[hombit]

We should highlight that from_dataframe() is not designed for large datasets (>~1M rows) and that hipscat-import should be used instead. This should go to both docstring and RTD

[smcguire-cmu]

This would also be good as a warning when we run from_dataframe to check the size of the input data frame and raise a warning if it's large.

[camposandro]

This has been completed by #379.