Improve protein parsing

[swooster]

@Ravna ran into an issue where invalid data wasn't caught early, resulting in much head-scratching: Two fastas had been concatenated without a newline between them, but FastaParser::<ProteinSequence> didn't catch that. This fixes that, as well as the fact that proteins don't strip out spaces, even though the tests imply that they should.

[swooster]

Important note: I wasn't sure whether to accept - as a valid protein character (wikipedia says that's commonly used to mean a gap of indeterminate length). I've opted to omit it for now.

[hwchen]

Since we're examining AA parsing, should we consider using an include-list for valid amino acids, instead of just checking ascii alpha?

I feel like there was a comment somewhere about just using u8 for now instead of having a proper AA type, but I don't think that needs to preclude us from parsing more strictly.

[swooster]

The reason I didn't include an AA list is because every single letter is some kind of AA code. Some are ambiguous and some are uncommon, but all letters are potentially valid.