Improve augur subsampling

[huddlej]

Logic to subsample sequences by metadata and other annotations has become more important and complicated. This logic currently lives in augur filter but we have considered a separate augur subsample command and other approaches to make subsampling faster, easier to add features to, and easier to maintain.

The following flow charts show examples of how an augur index command that generates summary statistics for sequence data could allow either the current augur filter or a new augur subsample command to speed up subsampling.

[huddlej]

This issue was partially resolved by the addition of the augur index command in #651 and is otherwise redundant with #635.