comments: cut the development log from the four largest offenders (#684)

Continues #683's standard — write it the way you would if you were porting the file to another language.
All four are TOKEN-IDENTICAL to their originals; comments only.

  slovenian/normalize.ts    458 → 380   (67-line header → 23)
  core/normalizeSymbols.ts  404 → 345   (185 of the remainder is genuine API spec)
  hungarian/normalize.ts    361 → 239   (107-line header → 35, the largest block in the repo)
  khmer/normalize.ts        307 → 236   (52-line header → 21)

Out: corpus tabulations of pre-fix readings, post-change measurements, discovery attribution ("reported
by the Ukrainian run"), and accounts of earlier wrong conclusions. In: every rule and trap — Slovenian's
ordinal-period discriminator and the no-utterance-final-pause-lost invariant, its four-way DUAL agreement
and why slavicCountForm cannot serve it, Hungarian's "lowercase follows" being stronger than "anything
follows", Khmer native digits outnumbering ASCII, and normalizeSymbols' longest-first ordering, trailing
apostrophe guard, mark guard and NOT_VERSION halves.

normalizeSymbols deliberately stopped at 345: 185 of those lines document the ~30 SymbolData fields, which
is specification a caller needs rather than narrative.

Found a real trap doing it — `*tri*/*štiri*` in a block comment forms `*/` and terminates it early.
Documented in CONTRIBUTING. A detector was written and then removed: it cannot be made precise (it also
flags legitimate `text*/` endings and glob strings like "src/**/*.ts"), and typecheck catches the real
thing every time since the remaining prose never parses.

Running total across both PRs: 146 → 142 files, 16,610 → ~14,300 comment lines over 1:1.
Prettier debt down again, 448 → 444. 224 test files / 3114 tests, typecheck clean.

Author: christopherthompson81

CONTRIBUTING.md

@@ -54,6 +54,11 @@ porter has no history to tell, so they write what the code does and what will bi
 
 Reference points in the tree: `src/core/hostWord.ts`, `src/languages/maori/normalize.ts`.
 
+⚠ Beware `*/` forming by accident inside a block comment. Markdown-style emphasis around a slashed
+alternative does it — `*tri*/*štiri*` terminates the comment early and the rest of the block parses as
+code. `npm run typecheck` always catches it (the remaining prose never parses), but the error points
+*past* the real cause and usually reads as an unterminated regex. Look for emphasis before a slash.
+
 When you cut comments, **do not touch code in the same commit**. Both files above were verified
 token-identical to their originals; that check is what makes a large comment pass reviewable.