fix(web): F3 defer parked tabs' cache replay to first activation (WI-127)

[thedancingdeveloper]

Third PR of the terminal-attach-budget initiative (WI-121). Stacked on #668 (F1) so the diff is F3-only; all land on feat/terminal-attach-budget.

The bug (WI-127)

On reload, every retained terminal tab replayed its cached tail (≤ 1 MiB each) immediately, through the single global replay FIFO, round-robin with the active pane. With N retained tabs the visible pane got 1/N of the parser — the "reload is slow" half of the terminal budget bug — then pre-warm cold-attached up to 8 more sessions on top.

The fix

A parked pane (retained-but-not-active, or the unfocused half of a split) now keeps its prepared cache tail in memory and replays it lazily on first activation (resumeSocket, before connect()), instead of scheduling into the shared FIFO on load. The active pane still replays immediately, so its restore is no longer time-sliced against tabs the user cannot see. The cache position is adopted at load time, so a warm reattach resumes correctly even if a re-park cancels the deferred replay before it runs.

The decision is a pure helper shouldDeferCacheReplay(parked, hasCachedBytes), unit-tested here. The six-cached-tabs end-to-end assertion (active pane replay unaffected by other tabs) lands with H3's mocked Playwright spec (WI-124) — that is F3's stated acceptance.

Verification

• pnpm typecheck clean; pnpm vitest — 909 passed (incl. new shouldDeferCacheReplay.test.ts).

🤖 Generated with Claude Code

https://claude.ai/code/session_01UtEFLJAhLiq4NfZCN2vuBb

[thedancingdeveloper]

Cleanup: superseded by the single promotion PR #676. All of this work is already integrated into feat/terminal-attach-budget and is reviewable there per-commit (each commit is one initiative item with a full message). Closing to keep the initiative to one PR; the commit is preserved.