Speed up large displayedRegions sets by converting MST array of Region[] into a types.frozen<IRegion[]>

[cmdcolin]

Pros:

• can handle very large sets of displayedRegions, which can occur from "show all regions" operations from a linear genome view, or showing entire genome in synteny view

Cons:

• changes the 'type spec' of the linear genome view, so calling getSnapshot on the displayedRegions object (or individual regions) is a thrown error (kind of annoying)
• can't directly "push" or mutate the self.displayedRegions array since it is not an observable after this
• there are problems with creating gigantic displayedRegions sets anyways because it causes the sessions to be too large to share, but that is sort of an aside

You can see benefit by performing a show all regions on our 'many_contigs' test dataset

main https://jbrowse.org/code/jb2/main/?config=test_data%2Fmany_contigs%2Fconfig.json

this branch https://jbrowse.org/code/jb2/large_displayed_regions_set/?config=test_data%2Fmany_contigs%2Fconfig.json

[cmdcolin]

before: 30+ seconds to load after a show all regions after: almost instant with somewhat slow side scrolling

[cmdcolin]

might go ahead with this one...could potentially cause small breakages (for getSnapshot of displayedRegions specifically but code can work around that if needed)