Implement hisat2 support

[roryk]

Via @schelhorn,

An additional comment: from my inspection of the available new aligners it seems to me as if going with the new hisat2 may be the best option for an evolution of the existing rnaseq pipeline. It's graph-based, includes known SNPs/indels/SVs in the index, and is supposed to be very fast, low-mem and accurate. Also, it is applicable to both WES/WGS and WTS libraries (since it knows about spliced alignment) and could even be used for variant2 and probably be better than bwa-mem due to the graph-based index. Last, it plays nice with cufflinks and stringtie. Since it outputs SAM it is a general-purpose tool (usable for assembly, disambiguation and fusion detection), which of course salmon and kallisto are not. Given all these advantages I'd suggest that it would make an excellent choice for a rnaseq2 prototype.

Agree on all fronts.