Fix CRAM mismatches calculation regression in v2.2.0

[cmdcolin]

The PR at #3293 introduced a regression where CRAM mismatches would be calculated incorrectly with the 'i' readFeature type due to incorrectly assuming that readFeatures would have an indication of stretches of matches. This PR fixes that assumption using code similar to the CIGAR generation

[codecov]

Codecov Report

Merging #3342 (85b6117) into main (8a7c3bf) will decrease coverage by 0.02%.
The diff coverage is 77.68%.

❗ Current head 85b6117 differs from pull request most recent head f18ed93. Consider uploading reports for the commit f18ed93 to get more accurate results

@@            Coverage Diff             @@
##             main    #3342      +/-   ##
==========================================
- Coverage   59.16%   59.14%   -0.03%     
==========================================
  Files         750      751       +1     
  Lines       29266    29270       +4     
  Branches     7084     7083       -1     
==========================================
- Hits        17316    17312       -4     
- Misses      11756    11764       +8     
  Partials      194      194              

Impacted Files	Coverage Δ
plugins/alignments/src/CramAdapter/util.ts	76.92% <76.92%> (ø)
...gnments/src/CramAdapter/CramSlightlyLazyFeature.ts	85.71% <100.00%> (+5.32%)	⬆️
...rative-view/src/ServerSideRenderedBlockContent.tsx	62.50% <0.00%> (-4.17%)	⬇️
...ns/alignments/src/PileupRenderer/PileupRenderer.ts	49.49% <0.00%> (-1.42%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[cmdcolin]

now does a "cross-check" by comparing the CIGAR and mismatches data structures from the same data file encoded in BAM and CRAM

this picked up a number of subtle logic bugs that were good to catch