restoring old (global) coverage calculation

[tyleraland]

Restore old coverage calculation.

Notes:

• This leaves "qcovs", as provided by blast, unchanged
• If args.coverage is provided, this calculates coverage internally and filters lines that don't meet the threshold. The internally calculated coverage isn't reported. This approach seems less disingenuous than clobbering blast's "qocvs" field.
• I opted to leave fields "qstart", "qend", "qlen" as implicit requirements so the code will fail noisily if they are omitted. Should we mention this in the argparse help text?

[nhoffman]

@tyleraland - I'd actually prefer to completely ignore qcovs since it is not particularly useful for our purposes, adds confusion (as we have seen), and will not necessarily be provided or calculated in the same way by other aligners (eg, vsearch). We should use the calculated coverage both for filtering and as the value included in the report.