Ordinary sentence commas dropped (unrelated to thousands separators)

Part of #5. Split off from #15's investigation.

#15 flagged that its own raw count of missing commas included "a smaller
number of ordinary sentence commas dropped for unrelated reasons." Re-running
`semcor-verify-brown`'s whole-document alignment against `nltk.corpus.brown`,
isolating pure single-comma deletions that are *not* between two digits,
confirms **52** such instances, spread across mixed genres (fiction,
learned, press) -- list-item commas, appositive commas, a comma dropped
next to a quote.

Unlike #15's thousands-separator case, there's no single shared trigger
across the sample -- each looks like an independent dropped comma rather
than one systematic transformation. Worth its own triage rather than
folding into #15's fix.

A further 3 instances are a related but distinct anomaly: a whole
`,` sequence (comma *and* digit) missing entirely, e.g. Brown has a
trailing `3,`/`7,`/`4,` that's fully absent from this corpus's `text` --
too rare to investigate on their own, but worth keeping in mind while
triaging the 52 above in case they share a cause.

### Suggested fix

No single mechanical fix suggests itself the way #15's did -- these likely
need per-instance triage against Brown context to confirm each is really a
missing comma (vs. e.g. a legitimate editorial choice) before fixing.

Start by rerunning `semcor-verify-brown`'s whole-document alignment against `nltk.corpus.brown`, isolating single-comma deletions that are not between digits. Review each of the 52 instances against Brown context, while noting the three missing digit-comma sequences; done means the cases are confirmed as errors or legitimate editorial choices and appropriate corrections are identified.

Written by the indexing model from the issue text.