Remove empty *features* from the visualization?

[fedarko]

This will probably come up eventually?

I imagine this will probably happen infrequently, but may be worth addressing. Similarly to removing empty samples, this would be done with the understanding that if all (or in this case two, I guess?) features are empty an error would be thrown.

I think that once #58 is taken care of (and _df_utils.remove_empty_samples() operates on BIOM tables), it should be possible to do this by just uh removing the axis="sample" argument to biom.Table.remove_empty(). We'd need to add some checks to make sure at least two features remain in the table post-filtering (just an extension of the current logic but on the observation axis), and we'd also need to add some logic to notify the user about dropped features (as well as document this behavior somewhere).

• Implement this in remove_empty_samples() (and rename that function to just remove_empty() or something).
• Notify the user about number of dropped features (analogous to current notifications re: dropped samples)
• Throw an error when all features are empty
• Add unit tests that this works
• Document this

[fedarko]

Just need to reconcile this with the test system (Byrd example is causing it to fail right now), then this will be done.