refactor(handlers): funnel HandlerRegistry's two maps through one insert/evict pair

[zzal]

Summary

Closes Part I Wave 3 #3 of the arch+DX audit. Every handler lived in both the instance handlers dict and the static globalTable, hand-synced across four methods (register, remove, closeScope, deinit). A missed write in any of them leaked a globalTable entry (a handler outliving its component) or corrupted scope diagnostics.

Why not the audit's "one authoritative map"

Both maps are genuinely load-bearing for different consumers, so they can't be collapsed:

• globalTable (static, all registries) backs dispatchGlobal — used by the single window.__swiflowDispatch JS callback, which receives only an integer handler ID and can't know which registry owns it.
• The instance handlers dict backs dispatch — used by TestRenderer, per-instance so parallel host tests don't cross-contaminate through a shared static.

Approach — one mutation funnel

Instead of collapsing the maps, funnel all mutation through one private pair:

private func insert(_ handler: EventHandler, scope: ScopeID?)  // the ONE add path
private func evict(_ id: Int)                                  // the ONE remove path

register / remove / closeScope / deinit now route through these, so the two maps (and a handler's scope attribution) can never drift — one add path, one remove path. Every one of the four call sites got shorter.

Test plan

• 4 new sync-invariant tests pinning the property the funnel guarantees: register → both dispatch paths reach the handler; remove/closeScope → neither does; a released registry (deinit) drops its handlers from the global table (no leak).
• All existing HandlerRegistry tests pass unchanged — the refactor is behavior-preserving.
• Full Swift monorepo suite 1361/273 (+4).
• Counter Playwright suite 11/11 — real-browser button clicks flow through dispatchGlobal → the refactored globalTable, exercising the production dispatch path end-to-end.

🤖 Generated with Claude Code

[github-actions]

📦 Bundle size

Artifact	Baseline	This PR	Δ
App.wasm	5.06 MB	5.15 MB	+1.88%
App.wasm (gzip)	1.80 MB	1.84 MB	+2.18%
JS runtime	54.5 KB	54.5 KB	+0.00%
JS runtime (gzip)	11.3 KB	11.3 KB	+0.00%
Total (gzip)	1.81 MB	1.85 MB	+2.16%

✅ Within budget (≤5% growth allowed).

_{Baseline: Swift 6.3, WASM SDK 6.3-RELEASE, measured 2026-06-18.}