feat: export script names and analyzer constants

[chrisdengso]

Callers writing rules against dominantScript or expectedScripts had no way to discover the actual values — they had to hardcode strings and hope. This exports what the analysis already knows.

New exports

Export	What it is
SCRIPT_NAMES	Every script the classifier can name — the possible values of dominantScript and words[].scripts
SUPPORTED_SCRIPTS	The subset this runtime's Unicode tables support (all of them, on a current engine)
PSEUDO_SCRIPTS	Common, Inherited, Unknown — never appear in a finding
FORMAT_CHAR_SCRIPTS	Scripts exempt from invisible/zalgo
LEGITIMATE_SCRIPT_COMBINATIONS	Mixes that never fire mixed_script
ZALGO_MARK_RUN	Marks per base at which zalgo fires
primaryScript(char)	Classify a single character
ScriptName, PseudoScript	The types behind them

Type narrowing

AnalysisResult.dominantScript, WordFinding.scripts and AnalyzeOptions.expectedScripts move from string to ScriptName. Runtime behavior is unchanged, but TS callers passing a plain string[] to expectedScripts now need ScriptName[] — which is the point: expectedScripts: ['Cyrilic'] no longer compiles silently matching nothing.

Testing

test/exports.test.ts ties each constant back to the behavior it claims to describe — ZALGO_MARK_RUN is asserted as the depth at which the signal flips, FORMAT_CHAR_SCRIPTS against a Persian ZWNJ passing and a Latin one failing — so a constant cannot drift from the analyzer without a red test. The README example runs in test/readme.test.ts per repo convention.

275 tests pass; typecheck, lint and build clean. Changeset included (minor).

🤖 Generated with Claude Code