[AZ-895] Deprecate replay auto-sync surface; file AZ-908 follow-up

Option A (minimum-deprecation, 2 SP) per user complexity-budget
decision. Auto-sync stays importable as a raising stub for one cycle
so external callers see a clean ReplayInputAdapterError instead of an
ImportError. Full physical removal is filed as AZ-908 (cycle-5+ backlog).

Production:
- auto_sync.py: 700+ LOC -> 56-line no-op stub raising
  "auto-sync removed; supply --imu CSV instead"
- tlog_video_adapter.py: 700+ LOC -> 105-line deprecated stub;
  ReplayInputAdapter.open() raises immediately, close() is a no-op
- _replay_branch.py: dropped legacy auto-sync branch +
  _build_auto_sync_config; _validate_replay_paths now requires
  imu_csv_path; replay_input_adapter_factory parameter removed
- cli/replay.py: --time-offset-ms / --skip-auto-sync / --auto-trim
  emit DeprecationWarning + stderr line; values ignored
- tlog_replay_adapter.py + tlog_ground_truth.py docstrings: AUDIT-ONLY

Tests:
- DELETED test_az405_auto_sync, test_az405_replay_input_adapter,
  test_az698_window_alignment (covered code no longer runs)
- ADDED test_az895_auto_sync_deprecated_stub (5 parametrised, pins AC-1)
- test_az402_replay_cli: deprecation warnings + ignored-value asserts
- test_az401_compose_root_replay: new imu_csv_path-required gate;
  deleted the calibration-loading test that relied on the removed
  replay_input_adapter_factory injection point
- test_derkachi_real_tlog: xfail reason refreshed to AZ-848 + AZ-883
  (AC-4 "AZ-848-scoped reason")

Docs:
- module-layout.md: replay_input file list flags deprecated modules,
  adds csv_ground_truth.py
- _dependencies_table.md: +AZ-908 row, preamble + totals updated
  (179 -> 180 tasks, 567 -> 570 SP)
- AZ-908 backlog spec added; AZ-895 spec moved todo -> done
- batch_03_cycle4_report.md written

Touched-module tests green (111 passed, 1 skipped). Full unit suite
green: 2287 passed, 85 skipped, 1 deselected (pre-existing flaky perf
test, unrelated).

Co-authored-by: Cursor <cursoragent@cursor.com>

tests/e2e/replay/test_derkachi_real_tlog.py

@@ -173,25 +173,20 @@ def _load_full_ground_truth(tlog_path: Path) -> list[GroundTruthRow]:
 @pytest.mark.tier2
 @pytest.mark.xfail(
     reason=(
-        "Blocked by AZ-776 + AZ-777. AZ-699 was implemented without "
-        "executing this test end-to-end on Tier-2 Jetson; once the "
-        "fixtures (real video + factory calibration) landed and the "
-        "test ran for real, two upstream gaps surfaced: (1) AZ-776 "
-        "— c4_pose ISam2GraphHandle Protocol rejects the ESKF stub "
-        "handle, so the c5_state=eskf composition variant cannot run; "
-        "(2) AZ-777 — Derkachi has no C6 reference tile cache / "
-        "descriptor index, so the default c5_state=gtsam_isam2 "
-        "composition reaches the per-frame loop but iSAM2.update "
-        "fails at frame 1 with key 'x2' not in Values (no C4 anchor "
-        "was ever inserted because C2/C3/C4 have nothing to match "
-        "against). Per AZ-777 AC-4: 'After AZ-776 + this ticket "
-        "both ship, test_ac3_within_100m_80pct_of_ticks can be "
-        "un-xfail'd and pass'. The AZ-699 verdict-on-real-flight is "
-        "tracked under those tickets; this xfail is the documented "
-        "mask until they ship. NOTE: this contradicts AZ-699 AC-1 "
-        "('no @xfail mask'); the dependency gap was discovered "
-        "post-implementation when the Jetson e2e harness ran for "
-        "the first time."
+        "Blocked by AZ-848 (+ AZ-883). The tlog adapter path is "
+        "structurally broken at the clock layer: VioOutput.emitted_at_ns "
+        "is sourced from process monotonic_ns (klt_ransac.py:274) while "
+        "ImuWindow.ts_end_ns comes from the FC IMU timebase, so the C5 "
+        "ESKF immediately diverges (mahalanobis² > 100) at frame 3 "
+        "with c5.state.eskf_out_of_order. AZ-883 is the related "
+        "SCALED_IMU2.ts_ns=0 default that exacerbates the mismatch on "
+        "some tlogs. AZ-895 made the (video, CSV) path the primary "
+        "replay surface, so AZ-848 is no longer bench-blocking — but "
+        "this test still exercises the legacy tlog path and will "
+        "remain xfail until AZ-848 / AZ-883 ship. The AZ-776 / AZ-777 "
+        "gaps that originally surfaced via this test were closed in "
+        "cycle 3, but the clock-mismatch root cause then surfaced "
+        "underneath."
     ),
     strict=False,
 )