BUG: updated dataframe sorting for identical values

[lizgehret]

this addresses test failures in the latest amplicon distro trial that includes some dependency updates: qiime2/distributions#1162

the issue was two-fold:

1. differing level of decimal place accuracy in the importances (at values very close to zero)
2. differing sort orders for identical values (primarily the list of zeros in importances)

Tests have been adjusted with a tolerance and importances are now doubly sorted - first on importance values, then by the index.

[lizgehret]

ci failures for osx are because there isn't a built prepare for osx with the new updates. here is the paired LANG PR that should be followed to confirm this is passing:

qiime2/distributions#1178

[ebolyen]

This is great! The actual values never changed and now the index order for zeros isn't arbitrary so we can resist future problems there.

Also it happens to make the existing sorts better via the updated util so the code became simpler.