Deal with multi-significant matches

[rdfinn]

Need to think about how to deal with sequences that contain multiple matches.

Issues
Multiple independent matches:
e.g. http://www.uniprot.org/uniprot/A0A101BN06 has multiple matches to PTHR45527.

Multiple local matches:
http://www.uniprot.org/uniprot/B8E2P5, where there are two local matches against model PTHR10000.

• Note both could be affected by minor alignment overlaps.

Multiple matches (either case) to independent PANTHER models:
Then there is how to handle gene fusion, where two different PANTHER matches do not get handled.

The first example will currently break tree grafting hmmer parsing/alignment code - well the length will get too long.