perf(dom): List.each bulk first-emission mount fast path (#178)

[stefvw93]

Closes #178.

What

A List.each region's first emission has no previous state, yet it paid the full keyed-reconciliation machinery designed for updates: per-item projection into an immutable HashSet (dup guard) and HashMap (identity map), a prevIndex build, a drop-set diff, and a longestIncreasingSubsequence pass over an all--1 sources array. For large static lists (e.g. 10k rows), first render was materially slower than the same items as plain array children.

reconcileList now takes a bulk mount fast path when previous state is empty:

• Trigger: prev.order.length === 0 && HashMap.isEmpty(prev.records).
• Reached by renderList's fresh commits, empty→refill, and hydrateList's HY2 divergence recovery (all pass an empty prev).
• The Effect-Equal-aware duplicate guard (projectKeys, KR1) and per-item renderItem (MR2, per-item scope) run exactly as the general path; then all item ranges insert in one pass before the region end marker and the identity map is built via HashMap.fromIterable.
• Skips prevIndex, the drop-set walk, and LIS (all no-ops against empty state).

The returned ListState is identical to the general path, so every later reconcile behaves the same. This is a behavior-preserving optimization; no public API changed.

Why it's equivalent

For all-new items the general path's LIS keep-set is empty, so its right-to-left anchored insert produces the same final DOM order as the fast path's single-pass insert. Renders happen before any insertion in both paths, so a render failure inserts nothing. HashMap.fromIterable yields a lookup-equivalent identity map (keys already de-duplicated).

Tests

• list-mount-fast-path.test.ts (jsdom): FE6 empty→refill re-entry, FE1 bulk mount at scale, FE2 record-state parity (reorder + combined insert/remove/reorder after a fast-path mount).
• perf-list-mount.browser.test.ts (Chromium): FE7 permanent guard, 10k static rows, structural assertions primary (10k li + 10k markers + order), same-run plain-children baseline + generous timing caps.
• The existing list.test.ts suite already exercises the empty-prev branch (it mounts from a non-empty SubscriptionRef), and list.hydrate.test.ts's HY2 divergence test covers the hydration-divergence entry.

Validation

All four gates green locally: vp run check, vp run test (799), vp run test:browser (70), vp run test:types (149 tests / 251 assertions).

Docs

docs/how-to/render-keyed-lists.md: new "Static lists: use plain children" section and a strengthened by: index note. list.specs.md amendment + new list-mount-fast-path.specs.md (ACs FE1–FE7).

🤖 Generated with Claude Code