Canadian French (fr-CA) as its own locale, not a fallback

[distronode-com]

First of the features split out of #30, in the order of appetite you described on that thread.

fr-CA currently resolves to the France copy through the primary-subtag fallback. The differences are not stylistic:

	fr	fr-CA
email	e-mail	courriel
reschedule	reprogrammer / reprogrammation	reporter / report
personal data	données personnelles	renseignements personnels
before ! ? ;	narrow no-break space	no space
July, abbreviated	juil.	juill.

The third row is the one that matters most: renseignements personnels is the statutory term in Quebec, so the France word is wrong in a legal sense rather than a register one. The last row is CLDR's own disagreement, not a preference, which is why TestDateTablesMatchCLDR catches it.

The only structural point

This is the first regional locale. Resolve already prefers an exact tag before falling back to the primary subtag, so no matcher change was needed — but that means the risk is not in the new locale, it is in the eight that already work. A visitor sending fr-FR or plain fr must be untouched by this file existing.

TestResolve now pins both directions:

{"Canadian French resolves to its own locale", "fr-CA,fr;q=0.9", "", "fr-CA"},
{"European French is unaffected by fr-CA", "fr-FR,fr;q=0.9", "", "fr"},
{"plain fr is unaffected by fr-CA", "fr", "", "fr"},
{"fr-CA can be selected by the ?lang override", "en", "fr-CA", "fr-CA"},

Mutation-checked rather than assumed: with the locale file moved aside, the two positive cases fail (Resolve("fr-CA,fr;q=0.9", "") = "fr", want "fr-CA") while the two fr/fr-FR cases stay green, which is what they are for — they assert an absence of change and would be worthless if removing the feature moved them.

What is verified and what is not

The three guards the other eight locales pass apply unchanged: same key set as English (TestAllLocalesHaveTheSameKeys), matching printf verbs including the %[1]s/%[2]s positional pair no_available_times_host needs (TestAllLocalesHaveMatchingFormatVerbs), and date tables cross-checked against CLDR (TestDateTablesMatchCLDR). So the structure is verified.

⚠️ The wording is not. It is an unreviewed draft, like every non-English locale in this repository, and nobody who speaks Canadian French has read it. The CHANGELOG says so plainly, so a correction arrives as an ordinary pull request rather than as a bug report from a Quebec booker.

Currency and percent take a non-breaking space before $ and % in Canadian French. No key carries either today, so that rule is recorded in ARCHITECTURE §23 rather than applied — worth writing down before someone adds a priced-event string.

Scope

Nine files: the locale, four test cases, and the doc/CHANGELOG counts that say "8 locales". No handler, no matcher, no schema. internal/i18n passes; the full suite passes on this branch.

Adding a locale is still just adding internal/i18n/locales/<code>.json — init() globs the directory, and nothing else needed a list updating.

🤖 Generated with Claude Code

[pullfrog]

ℹ️ No critical issues — two stale locale counts outside the files this PR already updated.

Reviewed changes One commit adding Canadian French as a first-class regional locale (data + Resolve pins + doc counts). No matcher or handler changes.

• fr-CA.json locale table New 158-key file with the claimed fr vs fr-CA differences (courriel, reporter/report, renseignements personnels, punctuation spacing, CLDR juill.). Key set matches en; package tests pass.
• TestResolve pins both directions Exact fr-CA, European fr-FR/fr unaffected, and ?lang=fr-CA override — the half that would break silently if the matcher preferred the new regional tag too broadly.
• Doc/CHANGELOG counts CLAUDE.md, CONTRIBUTING.md, ARCHITECTURE §23, and the Unreleased CHANGELOG entry now say 9 locales / list fr-CA.

ℹ️ Locale counts still say 8 in two places this PR did not touch

README.md still markets "8 languages" and lists English through Swedish with no Canadian French — the most visible user-facing surface for the feature. docs/ARCHITECTURE.md (~L809, LiveKit "not translated" known issue) still says the booking surfaces "ship in 8 languages (§23)", which now contradicts §23's "9 locales". Historical "8 locales" changelog entries are fine to leave.

Technical details

# Stale locale counts outside the PR's doc set

## Affected sites
- `README.md` ~L193–195 — still "**8 languages**" and the bullet list without Canadian French
- `docs/ARCHITECTURE.md` ~L808–809 — LiveKit known-issue still "ship in 8 languages (§23)" while §23 now says 9

## Required outcome
- User-facing and cross-referenced live counts match the shipped locale set (`en es fr fr-CA de it pt nl sv` / 9)
- Do not rewrite historical CHANGELOG / ARCHITECTURE dated entries that correctly described the state at ship time

## Suggested approach (optional)
- Bump README to 9 languages and add Canadian French to the list (same register as the other names)
- Change ARCHITECTURE L809 "8 languages" → "9 languages" (or "§23's locales") so it stays consistent with §23

  ^{｜ Fix it ➔ ｜ View workflow run ｜ Using Grok ｜ 𝕏}