parallelise parse_seqids

[oliverdrechsel]

parse_seqids is called multiple times and is a very slow process. on 57000 unique sequence headers it takes on our system more than 30min.

This might be avoidable, if

1. the for loop over all input files would parse multiple files simultaneously
2. wouldn't be called twice once for sequence names and later for sequence length

[pirovc]

You are right on those points, thanks for the suggestions @oliverdrechsel, will try to implement for the next release