Use the pivot table directly for taxa

[jonbrenas]

Addresses #1268

[adilraza99]

Hi @jonbrenas, really nice fix. Deriving taxa from the pivot table makes the data flow much more consistent and resolves the mismatch cleanly.

One small non-blocking thought: if we capture taxa from df_pivot.columns right after the pivot, before merging with df_location_aggs, we could skip the exclusion list and the int() cast. The merge is what introduces those non-taxa columns and the float dtype. Totally optional since the current approach works fine.

Also, would it be worth adding a small regression test for this scenario (samples with no location data or colony-only datasets)? Happy to put one together if that would help.

[jonbrenas]

I think you are right. I don't think the tests are needed, though.