Fix rendering of UCSC repeatmasker BigBed and BED files

[cmdcolin]

UCSC RepeatMasker BED files have extended fields which are somewhat complicated

This PR fixes the rendering of repeatmasker BED, BEDTabix, and BigBed files from UCSC with full fields

Example here https://hgdownload.soe.ucsc.edu/gbdb/hs1/t2tRepeatMasker/

See the "Full mode visualization" which shows, at least from the UCSC UI perspective, how complex these data are...https://genome.ucsc.edu/cgi-bin/hgTrackUi?hgsid=2372603109_AcR4Yt0XayFS0KE6hJ1sksSrAkTv&db=hub_3671779_hs1&c=chr1&g=hub_3671779_t2tRepeatMasker

In the data format, they re-use UCSC specific BED fields and specifically troublesome is they make negative valued blockSizes to indicate unaligned regions. This caused JBrowse to create negative length SimpleFeature's which crashes the track (we just don't allow negative size features, probably best to not change that)

This fixes it by filtering out the negative size features, and also does some heuristics to check if the track is a repeatmasker track to avoid it being inferred to be a "ProcessedTranscript" (it is hard to exactly tell since RepeatMasker uses thickStart,thickEnd, blockSizes, etc. this new heuristic is that description.split(' ').length == 15. hacky, but it is otherwise hard to tell)