🐛 Fix(e2e): resolve the Daily checkout without the cwd and run the AIDP journey on the mock only

[cj2026-bit]

Problem

The Ubuntu Daily run failed 36 seconds after the 2026-10-15 schedule. The launcher starts from the GitHub runner workspace (/home/jason/actions-runner/_work/...) while the checkout lives at /root/project/nexent; the external-knowledge preflight resolved the repository from the process working directory (repository_root()) and aborted the run with ValueError: Cannot locate repository root. The same change had also turned a per-case dependency into a batch gate: a missing machine-local profile config (config/aidp-journey.yaml) was reported as a doctor problem, and Daily refuses to start on any doctor problem — one optional case held the whole run hostage.

Fix

• Explicit checkout for the preflight: static_config_problems(home, records, *, repo) receives the checkout as an explicit argument, doctor passes its already-resolved checkout through, and nothing in the preflight resolves a repository from the process working directory any more. The launcher itself already resolves the checkout from its own file location, so no other production change is needed; the profile-config gate is gone.
• AIDP journey on the mock only: there is no real AIDP environment to smoke-test against (short and mid term), so the mock/real profile-selection mechanism is retired entirely. PW-KB-AIDP-01 declares its mock dependency directly (required_mock_services: [knowledge-provider]) like its sibling mock cases, the spec runs a single fixed seed oracle with no NEXENT_TEST_SELECTED_PROFILE / NEXENT_TEST_AIDP_REAL_* plumbing, ext_attach drops profile resolution and real_smoke export, the validator rejects profile_selection, and aidp-journey.example.yaml / schema are deleted. No machine needs a machine-local journey file at all.

Verification

Reproduced and verified end to end on a 1:1 WSL simulation of the Ubuntu Daily (launcher started from a runner-like working directory with the workflow's exact command, trimmed registry, the same machine-local build/deploy/preflight hooks):

• Old revision (bb4cdad0c): Daily exits 2 in ~1s and records only "ValueError"; a diagnostic re-run from the runner-like cwd reproduces Cannot locate repository root from .../runner-work/_work/nexent/nexent with the stack at local_preflight.py repository_root().
• Fixed revision (final HEAD of this branch): the Daily completes D0-D6 with the full related case set and --publish-latest green. Trimmed 4-case run PASSED 4/4 (warm 53s; cold 10m43s). Full related set (12 cases): 8 PASS (AGT-001/002, CTR-016/017/046/047, PERF-09, REL-02) — the four PW browser journeys fail inside the shared page-object locators (publish dialog and agent-editor buttons) that no longer match the current develop UI; those helpers are untouched by this PR and drive three cases it never modifies, so this is pre-existing suite/UI drift, not caused by this diff. PW-KB-AIDP-01's own mock chain — attach, real AIDP recall, oracle text and citation — passes preconditions 4/4 and steps 1-6, failing only at that shared publish-dialog drift.
• Offline: tools tests 437 passed with exactly the pre-existing 17-failure set (one existing onboarding test call site adapted to the new explicit-repo signature; no new cases added), asset validator green, copyright headers in place.

Operational note: a machine that wants the mock-backed cases to run needs the ext-knowledge mock deployed with its identity recorded in the machine-local config/ext-knowledge.yaml (deployment owns the mock lifecycle); without it those cases are dependency-blocked individually with a concrete receipt instead of affecting the batch.