default to minimap2 for assemble_refbased

[dpark01]

This PR includes commits for:

• bugfix to WDL code for handling aligner_options properly when aligner is something other than default/novoalign
• change default aligner to minimap2

A minimap2-based test run on 1500 SARS-CoV-2 genomes is in flight over here. Pilot test on 44 edge-case genomes (ie on the edge of being considered an acceptable assembly) shows that minimap2 is consistently faster (at least 2x) than licensed novoalign and always has an aligned read count greater than or equal to novoalign's (usually pretty close).

Assuming the test run looks consistent with previous novoalign-based outputs, this change has the benefit of

1. being accessible and fast to all users regardless of whether they have a novoalign license
2. extending assemble_refbased's default behavior to accept Oxford Nanopore and PacBio inputs in addition to Illumina based reads (although this has not yet been tested on real data). It can actually technically accept a mixture of reads created from different sequencers, as long as each BAM read group header is marked with the appropriate PL tag (whether or not that's actually a good idea is left to the discretion of the user).