properly merge semmedb sentences

[rjawesome]

biothings/biothings_explorer#833

[colleenXu]

@rjawesome

I tested this and Examples 1 and 2 now have 51 text-snippet edge attributes, rather than capping at 50. So....maybe that needs adjusting?

Also remember the additional ask in biothings/biothings_explorer#833 (comment)

[colleenXu]

@rjawesome @tokebe

Small thing, can it be done ASAP?
It looks like the evidence_count is being generated after truncating to 50 PMIDs, because both examples's evidence_count is exactly 50. If this is the case, is it possible to change this behavior so the evidence_count is the PMID count BEFORE truncation? I guess it's not a big deal, but it's an indicator to us that truncation is happening.

Otherwise it looks good: the off-by-1 problem is fixed.

(based on latest commits, described here)

[rjawesome]

@colleenXu the evidence count is fixed so it is counted before truncation