Optimization for SNPCoverageAdapter and CRAM parsing

[cmdcolin]

This can save about ~1second in generateCoverageBins calls on a 20kb region

Allocates less nested objects

On 200x coverage shortreads from jb2profile

Time spent in generateCoverageBins specifically:

This branch 2.82s
Main branch 3.6s

Possibly we could also bin multiple bp into a single bin (jb1 does this) and or use skips_and_dels (another jb1-ism) to optimize further, but we would need to be careful about binning multiple bp into a single bin with snps/modifications as those should probably not be aggregated

[codecov]

Codecov Report

Merging #2947 (9361a15) into main (d14ea8e) will decrease coverage by 0.06%.
The diff coverage is 62.65%.

@@            Coverage Diff             @@
##             main    #2947      +/-   ##
==========================================
- Coverage   60.74%   60.67%   -0.07%     
==========================================
  Files         601      601              
  Lines       27329    27341      +12     
  Branches     6643     6654      +11     
==========================================
- Hits        16600    16590      -10     
- Misses      10419    10441      +22     
  Partials      310      310              

Impacted Files	Coverage Δ
...rc/LinearSNPCoverageDisplay/components/Tooltip.tsx	29.62% <0.00%> (ø)
...ments/src/SNPCoverageAdapter/SNPCoverageAdapter.ts	56.37% <50.00%> (-2.81%)	⬇️
...s/alignments/src/PileupRenderer/PileupRenderer.tsx	48.20% <60.86%> (-0.38%)	⬇️
...gnments/src/CramAdapter/CramSlightlyLazyFeature.ts	80.20% <93.33%> (-5.44%)	⬇️
...ckages/core/pluggableElementTypes/RpcMethodType.ts	83.33% <0.00%> (-2.09%)	⬇️
...iew/src/LinearGenomeView/components/CenterLine.tsx	100.00% <0.00%> (+8.33%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d14ea8e...9361a15. Read the comment docs.

[cmdcolin]

This PR now also adds some CRAM optimizations too (raw loops instead of foreach, preallocate mismatches). Combined with @gmod/cram optimization, this branch is about ~20% faster than main (the @gmod/cram was about 10%-15% faster) for longread CRAM

400xcram longread
main 39.545
this branch 29.794
1.2x faster

600xcram longread
main 51.299s
this branch 39.545s
1.2x faster

800xcram longread
main 63.608s
this branch 51.496s
1.2x faster

[cmdcolin]

note: these changes above improve the longread performance more while the @gmod/cram pr improve shortread more