feat: improved FASTA parsing

[valentynbez]

Co-authored by: Fabian Kloetzl @kloetzl

• improves speed of FASTA file parsing to 2GBs, see improve parsing performance lh3/seqtk#123
• all regression tests passed

[milot-mirdita]

Thanks you! The change makes sense, however, we have actually modified this part. We check here if a sequence is broken across multiple lines or if its a single line sequence. This enables some different optimizations later in MMseqs2.

[kloetzl]

Yeah, the patch needs to be changed accordingly. Shouldn't be too hard to do.

[valentynbez]

Sorry, I missed this if statement completely. I see why it passed the regression test.