Streaming FASTA parser

[vgel]

Discussed in #9

Right now, the FASTA parser inhales the whole file and produces a Vec of records. For very large files, we might prefer a streaming API. Possible inspiration is https://github.com/raphlinus/pulldown-cmark

[vgel]

(also, once we have a streaming API, we can re-implement the Vec-based API in terms of it very easily)