Roadmap: lock §5a locale design; reflect shipped work

[mmucklo]

Converts the §5a "Open design questions" into resolved decisions so implementation of v2.1 (locale support + extension APIs) can start without another round-trip.

Key design decisions locked

• Locale contract — abstract class with protected instance rule tables + shared regex engine. Bare interface deferred.
• Visibility — protected rule tables; defaults come from protected const class constants on subclasses. No reintroduction of public static mutable state.
• Caching — per-instance, not global. Extension methods invalidate the instance cache.
• Default locale (static API) — always En. No `setDefaultLocale` footgun.
• Instance API — `new Inflect(Locale|string $locale = 'en')`; `Inflect::registerLocale(name, class-string|Locale)` for third parties; lazy resolution.
• Back-compat — static methods keep signatures, internally delegate to a lazy shared `En` instance. Proxy extension methods on `Inflect` mutate that shared instance.

Phasing tightened

• v2.0 ✅ shipped 2026-04-13.
• v2.1 — items 5 + 5a combined (they share the same API surface — splitting was churn).
• v2.2 — add at least one non-English locale (candidate: `Es`/`Fr`/`De`) to prove the contract holds.
• v3.x (conditional) — Path B package split per §5a triggers.

Ship-state annotations

• §6 docs: largely shipped in Docs: expand README, add Keep-a-Changelog CHANGELOG.md and CONTRIBUTING.md (roadmap §6) #19 (remaining: document instance API + locale extension once §5a lands).
• §7 tooling: phpstan level 8 + cs-fixer shipped in Add PHPStan level 8 + PHP-CS-Fixer; fix surfaced issues (roadmap §7) #18; infection + phpbench deferred.
• §8: v2.0.0 tagged.

Test plan

• Confirm the Locale contract shape (abstract class vs interface, protected visibility, constructor-seeding).
• Confirm the no-`setDefaultLocale` decision.
• Confirm v2.1 combining §5 and §5a is the right call.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.07%. Comparing base (00188a4) to head (b1da2ae).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##             master      #20   +/-   ##
=========================================
  Coverage     98.07%   98.07%           
  Complexity       27       27           
=========================================
  Files             1        1           
  Lines            52       52           
=========================================
  Hits             51       51           
  Misses            1        1           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.