[AZ-595] Batch 75: sitl_observer FDR-replay + scenario probe cleanup

Implement all 11 `sitl_observer` public surfaces as an offline
FDR-replay strategy (reads JSON fixtures under `${E2E_SITL_REPLAY_DIR}`
instead of live pymavlink/yamspy). Replace 12 per-scenario
`_harness_helpers_implemented` probes with one shared session-scoped
`sitl_replay_ready` fixture in `e2e/tests/conftest.py`.

Net: -636 LoC of duplicated scenario gating, +17 LoC shared fixture,
+38 new unit tests (596 total, up from 558). Includes K=3 cumulative
review for batches 73-75 (PASS).

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

_docs/02_tasks/done/AZ-595_sitl_observer_fdr_replay.md

@@ -0,0 +1,81 @@
+# sitl_observer FDR-replay surfaces + scenario probe cleanup
+
+**Task**: AZ-595_sitl_observer_fdr_replay
+**Name**: Implement all 11 sitl_observer surfaces with an FDR-replay strategy + refactor scenario probes
+**Description**: Replace `NotImplementedError` and stub patterns across `sitl_observer.py` with file-backed JSON readers. Refactor the brittle `_harness_helpers_implemented` probes in 8 scenarios to use an explicit `E2E_SITL_REPLAY_DIR` env-var check.
+**Complexity**: 5 points
+**Dependencies**: AZ-406, AZ-594
+**Component**: Blackbox Tests / Test Infrastructure (epic AZ-262)
+**Tracker**: AZ-595
+**Epic**: AZ-262 (E-BBT)
+
+## Problem
+
+The `sitl_observer` module reserved 11 surfaces in AZ-406 and never
+came back to fill them. Scenarios in batches 71-73 work around it
+with a `_harness_helpers_implemented` probe that passes a fake path
+to each helper and inspects the exception type — fragile and
+misleading post-fix once underlying helpers stop raising
+`NotImplementedError` (as they now do after batch 74).
+
+## Outcome
+
+- `sitl_observer.py` implements all 11 surfaces using an FDR-replay
+  strategy: each surface reads its fixture from
+  `${E2E_SITL_REPLAY_DIR}/<surface_name>.json` and returns a typed
+  result. Missing env var or missing file → empty result for `read_*`
+  surfaces; explicit `RuntimeError` for surfaces that require live
+  data.
+- Scenario probes refactored to use a single `_e2e_sitl_replay_dir_available`
+  fixture that returns True iff env var is set + directory exists.
+  No more fake-path exception-introspection.
+
+## Scope
+
+### Included
+- `sitl_observer.get_observer` + 10 free-function surfaces.
+- Typed dataclasses for the read_* / capture_* / observe_* / query_*
+  return types.
+- Comprehensive unit tests.
+- Probe refactor across the 8 affected scenarios.
+
+### Excluded
+- `fc_proxy_runtime` driver (separate ticket).
+- Live pymavlink / yamspy / TCP plumbing.
+- Per-scenario fixture builders (the JSON files that go in
+  `E2E_SITL_REPLAY_DIR` for each scenario).
+
+## Acceptance Criteria
+
+**AC-1**: Each surface returns its typed result by parsing JSON at
+`${E2E_SITL_REPLAY_DIR}/<surface_name>.json`. Missing env var →
+returns empty list / vacuous result (for `read_*`) OR raises
+`RuntimeError` (for surfaces requiring non-empty fixtures).
+
+**AC-2**: Probe pattern in the 8 affected scenarios replaced with
+`_e2e_sitl_replay_dir_available` fixture (True only when env var set
+AND directory exists).
+
+**AC-3**: ≥5 unit tests per surface category.
+
+**AC-4**: Full e2e unit-test suite passes (regression gate).
+
+## System Under Test Boundary
+
+None — runner-side helper that parses runner-produced fixture files.
+No `src/gps_denied_onboard` imports.
+
+## Constraints
+
+- Same `time.sleep` injection / `realtime` pattern as
+  `frame_source_replay` and `imu_replay` where pacing applies (e.g.
+  `capture_ap_tlog(duration_s)` should not block real wall-clock in
+  tests).
+- File-not-found surfaces with clear `RuntimeError` messages that
+  point at the env var.
+- All public dataclasses immutable (`frozen=True`).
+
+## Document Dependencies
+
+- `_docs/02_document/tests/blackbox-tests.md` § Test infrastructure
+- `_docs/02_tasks/done/AZ-594_harness_stubs_core_three.md`