Fixes negative tlen filtering

[LudvigOlsen]

Hi Rafeed,

When tagging the BAM file, only one read per fragment were written.

The issue was that the template length was used directly as fragment length. But half of the reads have negative template lengths, so it needs to be the absolute value instead.

There was also a hardcoded 51bp minimum length which would lead to errors if another minimum length is selected by the user. So I set that to the start_len instead.

Note that the changes from #2 is also in this. So you could either merge that first or only merge this one.

Best,
Ludvig

[LudvigOlsen]

Note: This issue with template lengths is also there in the coverage track command. Not sure whether you want to fix that or just use one read for coverage, but I'm seeing some extreme outliers due to very high GC tagged weights (>50).