fix(keywords): fail loudly when a scanner tier is unregistered or contradicts grammar.js

[dpezto]

Adding a scanner tier without updating SYMBOL_TIER has broken the dictionary generator twice — KW_FILTER_IF, then KW_G_ARGV — both times on the very change that introduced the tier. The existing unmapped-symbol assert did catch it, but only after mining, and it named the affected keyword (unmapped scanner symbol for by) rather than the file to edit.

SYMBOL_TIER stays hand-written rather than derived. The tier genuinely is in grammar.js — alias($.kw_g_argv, "arg") is exactly the mapping — but kw_dt and kw_lt each carry two aliases (dt/opt and lt/opt), so deriving would have to guess, and a guess writes a plausible dictionary with a wrong tier instead of failing. Convention-based derivation fails too: KW_G_ARGV must map to arg, not argv.

Three checks, all running before mining so they fire at the source rather than downstream:

• Missing symbol — every KW_* referenced by the scanner tables must appear in SYMBOL_TIER; the error names the file, the map's line, and the line to add.
• Wrong tier — for any symbol with exactly one alias in grammar.js, SYMBOL_TIER must agree with it. The alias is authoritative, so a contradicting hand-written entry is a silent dictionary defect that nothing previously caught.
• New multi-alias — the two deliberately multi-aliased tokens are listed explicitly rather than computed as "anything with more than one alias". A computed rule would silently absorb the next such token and weaken the check without a decision; instead a third one fails the build and forces the call. For the two listed, SYMBOL_TIER must still be one of that token's actual aliases, so a tier that is neither is caught.

All three failure modes were verified by deliberately breaking each in turn and confirming the message points at the fix. keywords.json regenerates byte-identical.

Runs inside the existing keywords CI job, whose first step is node scripts/gen-keywords.mjs — so these fire there, ahead of the git diff --exit-code keywords.json sync check.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!