Fix the operator settings split's stale callers

## Defect (observed on develop 61942ff65, after epic 7552's 21 split PRs merged) The `app/operator_settings.py` → `app/operator_settings/` package split (Task 7602, PR #6146) left callers that still reach typed readers on the bare package, which re-exports nothing (`app/operator_settings/__init__.py`, C1). 1. **Live runtime bug:** `apps/api/app/trigger_intake.py:52` does `from app import operator_settings`; `:191` `preparing_window()` calls `operator_settings.get_int("node_clone_timeout_secs")`. Repro: `cd apps/api && python3 -c "import app.operator_settings as o; print(hasattr(o,'get_int'))"` → `False`. `preparing_window()` is called from `trigger_intake.py:1388` for every PREPARING preflight row, so that path raises AttributeError in production. 2. **Stale test patches (develop gate red):** tests patch/read `<module>.operator_settings.get_*` on modules that now import `from app.operator_settings import reads as settings_reads`: - `apps/api/tests/test_cpu_profile.py:50` (`cpu_profile.operator_settings`) - `apps/api/tests/test_git_connect.py:262,282,375,418,421` (`git_access.operator_settings`) - `apps/api/tests/test_plan_matrix.py:98,100` (`plan_terms.operator_settings`) - `apps/api/tests/test_peer_lane_scheduling.py:123,187` (`fleet_routes.operator_settings`) — confirm against the module's current import. 3. Remaining gate-red items reported by draft Task 7718 while gating PR #6162: system-prompt goldens in `tests/project_memory/test_dispatch_plan_off.py` and `tests/test_project_memory_dispatch.py`; `tests/test_fleet_routes.py::test_the_quota_gate_is_read_for_the_asking_nodes_org`; jaahd `test_git_credential_helper` / `test_git_mirror_isolation`; a mypy arg-type violation in `apps/api/app/routes/cloud_nodes.py` above baseline. For each: confirm on develop whether it is split fallout; fix it if so, otherwise name it in the PR body with evidence that it fails on the pre-split base too. ## Fix (no new surface) - `trigger_intake.py`: import `from app.operator_settings import reads as settings_reads` and call `settings_reads.get_int(...)` (match the module's other operator-settings reads). - Each test: patch the reader on the name the module under test actually uses (`<module>.settings_reads`, attribute `get_*`), per Architecture C2. - Sweep: `git grep -nE '\boperator_settings\.(get_|_)' -- apps/ tools/` and fix every attribute access on the bare package. - Add one regression test that calls `trigger_intake.preparing_window()` and asserts it returns a `timedelta` (it raises today). - Draft Task 7718 covers the same scope; reference it in the PR body (it can be deleted once this merges). Out of scope: no product/behaviour changes; do not re-split or re-shape any package from epic 7552; do not add a shim/re-export to `app/operator_settings/__init__.py` (C1/N4); do not touch other packages' structure. Done means: - AC1 `python3 -c "import app.operator_settings as o; from app import trigger_intake; print(trigger_intake.preparing_window())"` (in the api container) prints a timedelta, no AttributeError; the new regression test passes. - AC2 `git grep -nE '\boperator_settings\.(get_|_)' -- apps/ tools/` → no attribute access on the bare package. - AC3 `./dc check full --gate-result --timeout 1500` → `GATE_RESULT status=pass` on the rebased tree. - AC4 Merged to develop through the merge lane; Jaah Deploy for the merge commit succeeds (cancelled/superseded → leave it). Required checks: ./dc check full

Fixed the operator_settings split's stale callers (7 test files plus a preparing_window regression test), shrank the mypy baseline, re-baselined goldens for Task 6722. Quick gate passed. PR #6166 queued in the merge lane. jaahd sandbox failures predate the split; filed as draft #7734.