v3.14.1

Patch. --parallel now actually parallelizes — inside real worktree lanes. The 3.14.0 flag silently degraded to sequential, and lanes shared the main working tree. Everything here came out of a real end-to-end run of the feature (live coder, 4-HU plan, --parallel 2).

Fixed

• --parallel N no longer degrades to sequential silently (KJC-BUG-0110): planToHuBatch dropped the plan's scope field during conversion, so the scheduler — which conservatively treats scopeless HUs as exclusive — never paired anything. Scope now survives into the batch (and through the plan-edit reconcile). Verified e2e: a 3-HU batch pairs disjoint-scope HUs into a real parallel chunk.
• Worktree lanes are now truly isolated (KJC-TSK-0629, PRs #1196-#1198): the coder prompt, pre-coder snapshot, acceptance tests, TDD/review diffs, sonar and the final commit/push all run INSIDE the lane's worktree on branch kj-hu-<id> — concurrent lanes never move the main working tree. Per-lane copies of config, pipeline flags, session and Brain state stop policies and reviewer feedback leaking across lanes. Also fixes a latent leak where max_iterations stayed clamped to hu_max_iterations after any acceptance-tested HU.
• kj plan ready no longer rejects every generated plan (KJC-BUG-0108): the auto-injected [PREFLIGHT-000] HU shipped without a status field and plan validation failed with "invalid status undefined".
• --yes answers questions that declare a safe default (KJC-BUG-0109): unattended runs stopped cold on the spec-review gate even at info severity with "(default: continue)" on screen. Callers can now declare defaultAnswer; --yes presses Enter for you and logs it. Questions without a declared default still stop loudly.