feat(review): dispatch wiring for e2eTests in queue/processors.ts

[JSONbored]

Part of #4189. Depends on #4192, #4193.

Context

Fix-handoff's dispatch shape in src/queue/processors.ts is a three-touch-point pattern, verified at the exact lines currently in main:

// processors.ts:8619
let fixHandoffEnabledForReview = false;

// processors.ts:9164
fixHandoffEnabledForReview = shouldEmitFixHandoff(env, repoFullName, reviewManifestForAutoReview?.review.fixHandoff ?? undefined);

// processors.ts:10592
...(fixHandoffEnabledForReview && aiReview?.inlineFindings?.length
  ? { fixHandoffBlocks: buildFixHandoffCollapsible(...) }
  : {}),

The critical invariant (called out explicitly in the surrounding comment at processors.ts:10589-10591): blocks are only built on a fresh cache-miss pass with inline findings present — a cache hit must render byte-identically, nothing new emitted. Getting this wrong means either regenerating (and re-billing) an LLM call on every cached re-render, or a stale block persisting after a repo turns the feature off.

Requirements

• Declare an outer-scoped e2eTestGenEnabledForReview = false flag near fix-handoff's own flag declaration (processors.ts:8619).
• Resolve it unconditionally alongside the other deterministic manifest overrides, using feat(review): resolver gate for e2eTests — src/review/e2e-test-gen.ts #4192's gate function against reviewManifestForAutoReview?.review.e2eTests (or the equivalent converged-feature accessor — confirm the exact manifest field name against feat(review): register e2eTests as the sixth converged-feature key #4190's schema work), mirroring processors.ts:9164.
• Build the actual blocks (calling feat(review): LLM core to turn a PR diff into Playwright E2E test source #4191's generation core, then feat(review): render + comment-assembly for generated E2E test blocks #4193's render function) only when e2eTestGenEnabledForReview is true, this is a fresh cache-miss pass, and aiReview?.inlineFindings?.length (or the equivalent "there's something worth generating tests for" condition) holds — mirroring processors.ts:10592-10594's exact cache-miss-only discipline.
• A cache-hit re-render must reuse the previously-generated block content rather than re-invoking the LLM — this feature is a real cost driver (an LLM call, potentially per push), so silent re-generation on every cached render is a correctness bug, not just a cost nit.
• No change needed to ENGINE_DECISION_GUARDRAIL_GLOBS (src/review/guardrail-config.ts:22-40) — sibling converged-feature wire files aren't individually listed there, and src/queue/** is already blanket-covered; this only changes if the feature later becomes a gate blocker (out of scope for this issue).

Deliverables

• The three processors.ts touch points wired, mirroring fix-handoff's exact shape.
• A test proving a cache-hit re-render does not re-invoke feat(review): LLM core to turn a PR diff into Playwright E2E test source #4191's LLM call.
• A test proving turning the feature off for a repo mid-flight removes the block on the next fresh pass (no stale content).

Expected outcome

Generated E2E test blocks appear in the PR comment exactly once per genuinely new review pass, never regenerated on a cache hit, and disappear cleanly the moment a repo's features.e2eTests flips to off or absent.

Resources / examples

• src/queue/processors.ts:8619, 9164, 10589-10594 (the exact fix-handoff touch points to mirror)

Effort

M — wiring is mechanical, but the cache-miss-only test is the part worth getting right.

[linear-code]

JSO-3226