docs: LADDER.md's framework prerequisites have all shipped

[Yaraslaut]

Closes #271.

Wider than reported

#271 reports that item 1 of LADDER.md's "Framework prerequisites" is stale.
Checking the other three found the same thing — all four have shipped:

#	prerequisite	shipped as
1	async shared/keyed attach for WASM	IBackend::registerModelSharedAsync (backend.hpp:187), attachModelAsync (:286), consumed by Bridge::ensureBoundAsync/attachHandlerAsync, implemented by QtWebSocketBackend
2	client-side execute deadline	Bridge::setExecuteDeadline (bridge.hpp:1099)
3	injectable time source	setNowOverride / ScopedNowOverride (util/datetime.hpp) — precisely the "process-global now-provider convention" the item asks for
4	fault-injection wire proxy	examples/common/testkit/fault_proxy.hpp, with strand_interleaver.hpp alongside it exactly as the item promised, both tested

The section opens with "Adversarial review found four items that invalidate
rung definitions-of-done as written" — so as it stood, it told a reader that
four things block the ladder when none of them do.

Kept, not deleted

The section explains why the rungs were sequenced as they were, which is still
worth knowing. What changes is its status: each item now says where it
landed, so a reader can check rather than take the document's word for it.

The "Also queued deliberately" paragraph

Same problem, but the three claims aged differently, so I didn't sweep them
together:

• offline queue depth bound — shipped (IOfflineQueue::maxDepth() +
  OfflineQueueFullError, reject-newest)
• SyncWorker's 5-attempt cap — still hard-coded at 5, so that half
  stands. What changed is that the surfacing hook rung 4 needs now exists
  (SyncWorker's DeadLetterSink parameter), making "surface dead-letters in
  the UI, not logs" buildable rather than blocked
• SQLITE_BUSY waits occupying pool threads — left untouched. It's a
  claim about runtime pool starvation, not about whether a symbol exists, and
  reading headers cannot settle it

Verification

Every path and symbol I wrote into the document was checked to exist first — 8
paths and 10 symbols. That felt worth doing explicitly here: writing an
unverified claim into a ladder document is the exact failure this PR is fixing.

Documentation only; scripts/check_spec_citations.sh passes.

Related

This is the fourth instance of the same defect class. #269 (TESTING.md +
polls/README.md: execute deadline and pendingCalls()) and #270
(polls/README.md: async attach) cover the other two files; both collide with
open PR #261 and are best done as one pass once it lands. LADDER.md is
collision-free, which is why this one goes first — and it is the document the
other two cite.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!