Make `filter_loom.py` much faster

[envest]

Try using loompy.combine along a shared axis (GRCh38 variant) to get everything in the same order, rather than loop through each variant