[URGENT] Redesign blast data loading

[lucventurini]

At the moment, I perform many of the calculations inside the serialization class (init), e.g. the calculation of the global identity.
This is MASSIVELY inefficient because I cannot perform bulk inserts the right way. I will have to preprocess the Blast Hits / Hsps BEFORE loading them into the database so that I can actually use the bulk_insert appropriately.

[lucventurini]

Probably solved by using raw __ table__.insert calls and pre-calculating the statistics before loading. I need access to the cluster data to confirm.

[lucventurini]

BLAST redesigned, unclear on the efficiency gains. This is now the preferred way, even inside the Hit class init.