fix(#1535): make punctuated adr-parser header synonyms reachable (#1536)

* fix: make punctuated CANONICAL_HEADERS synonyms reachable (audit M7)

classifyHeader received an already-normalized header (via normalizeAdrHeader,
which collapses [\s:._-]+ to a space and strips [^\w\s]) but compared it against
the RAW synonym strings. So any synonym carrying a hyphen/apostrophe ('trade-offs',
'non-goals', 'anti-goals', 'follow-up', 'cross-cuts', 'post-grilling', "how we'll
know", "won't do/have") could never match — its ADR section silently went unmapped.
Nine synonyms across six buckets were dead; the repo had characterization tests
pinning that quirk ('unreachable synonym').

Root-cause fix (per ADR-1372's 'compound, don't accrete' guidance): normalize BOTH
sides via a module-load-precomputed index, instead of pre-baking 9 normalized
literals into the data table. Closes the abstraction asymmetry once for all current
and future synonyms; the table stays human-readable. Also de-dupes 'trade-offs' from
considered_options so it no longer shadows risks once both normalize to 'trade offs'.

Insertion order preserved → first-match-wins + exact-then-prefix precedence byte-
identical for already-normalized synonyms; only the 9 dead synonyms gain matching.
Updates the 7 characterization tests to the corrected behavior and adds a
reachability+no-collision invariant test guarding the whole class against regression.

Note: the audit's M7 premise (trade-offs *misclassified into considered_options*)
was factually wrong — it was unmapped, and tested as such. adr-parser is CLI-only
(ADR-1372 T2), so the behavior change cannot reach in-process gates.

Claude-Session: https://claude.ai/code/session_01R88n7Q54bAaVHFkDbbH1yz

* chore(changeset): Fixed fragment for #1536 (adr-parser punctuated synonyms)

Claude-Session: https://claude.ai/code/session_01R88n7Q54bAaVHFkDbbH1yz

---------

Co-authored-by: Tom Boucher <trekkie@nomorestars.com>

src/adr-parser.cts

@@ -72,7 +72,6 @@ const CANONICAL_HEADERS: Record<CanonicalHeader, string[]> = {
     'candidates',
     'approaches considered',
     'variants',
-    'trade-offs',
     'pros and cons of the options',
     'discussion',
   ],
@@ -208,13 +207,30 @@ function normalizeAdrHeader(raw: unknown): string {
     .trim();
 }
 
-function classifyHeader(normalizedHeader: string): CanonicalHeader | null {
+// Normalized synonym index (audit M7). classifyHeader receives an ALREADY-normalized
+// header (via normalizeAdrHeader), but historically compared it against the RAW synonym
+// strings. Because normalizeAdrHeader collapses [\s:._-]+ to a space and strips [^\w\s],
+// any synonym carrying a hyphen/apostrophe/etc. ('trade-offs', "won't do", 'post-grilling')
+// could never match a normalized header — it was silently dead, and its ADR section went
+// unmapped. Normalizing BOTH sides closes that abstraction asymmetry once, so every synonym
+// (current and future) is reachable regardless of punctuation. Precomputed at module load to
+// avoid re-normalizing the whole table per call; insertion order is preserved so first-match-
+// wins and the exact-then-prefix precedence stay identical to the prior raw-compare loop.
+const _NORMALIZED_SYNONYM_INDEX: Array<[string, CanonicalHeader]> = (() => {
+  const index: Array<[string, CanonicalHeader]> = [];
   for (const [canonical, synonyms] of Object.entries(CANONICAL_HEADERS) as Array<[CanonicalHeader, string[]]>) {
     for (const synonym of synonyms) {
-      if (normalizedHeader === synonym) return canonical;
-      if (normalizedHeader.startsWith(`${synonym} `)) return canonical;
+      index.push([normalizeAdrHeader(synonym), canonical]);
     }
   }
+  return index;
+})();
+
+function classifyHeader(normalizedHeader: string): CanonicalHeader | null {
+  for (const [synonym, canonical] of _NORMALIZED_SYNONYM_INDEX) {
+    if (normalizedHeader === synonym) return canonical;
+    if (normalizedHeader.startsWith(`${synonym} `)) return canonical;
+  }
   return null;
 }