fix(core): centralize async/sync bridge for APScheduler scheduled jobs

Create AsyncJobRunner — centralized bridge between APScheduler
(BackgroundScheduler, sync thread pool) and async coroutines.

Fixes:
- P0: execute_run() called without await from APScheduler thread,
  causing coroutine to be silently discarded (root cause: no
  translation history)
- P0: get_async_job_runner() deadlock when called from APScheduler
  thread pool without running event loop
- P1: ID mismatch in disable_schedule/delete_schedule routes
  (job_id passed instead of schedule_id)
- P1: asyncio.run() in APScheduler callbacks incompatible with
  running event loop
- Delete unused llm_analysis/scheduler.py (not used in production)

Changes:
  core/async_job_runner.py          — new: AsyncJobRunner class
  core/scheduler.py                 — use runner.run()/run_later()
  translate/scheduler.py            — use runner.run() for execute_run
  mapping_service.py                — remove unused BackgroundScheduler
  dependencies.py                   — add get_async_job_runner() DI
  app.py                            — init runner in lifespan
  api/routes/migration.py           — use runner.run()
  _schedule_routes.py               — fix ID mismatch
  plugins/migration.py              — use runner.run()
  llm_analysis/scheduler.py         — delete (unused)
  tests: 151 new/updated tests, all passing

backend/tests/plugins/test_llm_analysis_scheduler.py

@@ -1,136 +0,0 @@
-# #region Test.LLMAnalysisScheduler [C:3] [TYPE Module] [SEMANTICS test, llm, analysis, scheduler, cron]
-# @BRIEF Tests for llm_analysis/scheduler.py — _parse_cron, schedule_dashboard_validation.
-# @RELATION BINDS_TO -> [LLMAnalysisScheduler]
-# @TEST_CONTRACT: _parse_cron -> dict | cron expression parsing
-# @TEST_CONTRACT: schedule_dashboard_validation -> None | schedules a validation job
-
-import sys
-from pathlib import Path
-sys.path.insert(0, str(Path(__file__).parent.parent.parent.parent / "src"))
-
-import os
-os.environ.setdefault("SECRET_KEY", "test-secret-key-for-testing")
-os.environ.setdefault("DATABASE_URL", "sqlite:///:memory:")
-os.environ.setdefault("AUTH_DATABASE_URL", "sqlite:///:memory:")
-
-from unittest.mock import AsyncMock, MagicMock, patch
-
-import pytest
-
-from src.plugins.llm_analysis.scheduler import _parse_cron, schedule_dashboard_validation
-
-
-class TestParseCron:
-    """_parse_cron — basic cron expression parser."""
-
-    def test_standard_5_part_cron(self):
-        """Standard 5-part cron expression parsed correctly."""
-        result = _parse_cron("0 8 * * 1-5")
-        assert result == {
-            "minute": "0",
-            "hour": "8",
-            "day": "*",
-            "month": "*",
-            "day_of_week": "1-5",
-        }
-
-    def test_daily_at_midnight(self):
-        """Daily at midnight."""
-        result = _parse_cron("0 0 * * *")
-        assert result["hour"] == "0"
-        assert result["minute"] == "0"
-
-    def test_every_hour(self):
-        """Every hour at minute 0."""
-        result = _parse_cron("0 * * * *")
-        assert result["minute"] == "0"
-        assert result["hour"] == "*"
-
-    def test_invalid_parts_count(self):
-        """Less or more than 5 parts returns empty dict."""
-        assert _parse_cron("0 8 * *") == {}
-        assert _parse_cron("0 8 * * 1-5 extra") == {}
-
-    def test_empty_string(self):
-        """Empty string returns empty dict."""
-        assert _parse_cron("") == {}
-
-    def test_step_values(self):
-        """Step values like */15 parsed correctly."""
-        result = _parse_cron("*/15 * * * *")
-        assert result["minute"] == "*/15"
-
-
-class TestScheduleDashboardValidation:
-    """schedule_dashboard_validation — schedules recurring validation."""
-
-    def test_schedule_with_params(self):
-        """Scheduler and task manager called with correct args."""
-        mock_scheduler = MagicMock()
-        mock_task_manager = MagicMock()
-
-        with patch("src.plugins.llm_analysis.scheduler.get_scheduler_service", return_value=mock_scheduler), \
-             patch("src.plugins.llm_analysis.scheduler.get_task_manager", return_value=mock_task_manager):
-
-            schedule_dashboard_validation(
-                dashboard_id="dash-1",
-                cron_expression="0 6 * * *",
-                params={"threshold": 0.8},
-            )
-
-        # Scheduler should have added a cron job
-        mock_scheduler.add_job.assert_called_once()
-        call_args = mock_scheduler.add_job.call_args
-        assert call_args[0][1] == "cron"  # trigger type
-        assert call_args[1]["id"] == "llm_val_dash-1"
-        assert call_args[1]["replace_existing"] is True
-        assert call_args[1]["minute"] == "0"
-        assert call_args[1]["hour"] == "6"
-
-    def test_schedule_creates_task_on_execution(self):
-        """The scheduled job function creates a task when executed."""
-        mock_scheduler = MagicMock()
-        mock_task_manager = AsyncMock()
-
-        with patch("src.plugins.llm_analysis.scheduler.get_scheduler_service", return_value=mock_scheduler), \
-             patch("src.plugins.llm_analysis.scheduler.get_task_manager", return_value=mock_task_manager):
-
-            schedule_dashboard_validation(
-                dashboard_id="dash-1",
-                cron_expression="0 6 * * *",
-                params={"threshold": 0.8},
-            )
-
-        # Get the job function and call it
-        job_func = mock_scheduler.add_job.call_args[0][0]
-        import asyncio
-        asyncio.run(job_func())
-
-        mock_task_manager.create_task.assert_called_once_with(
-            plugin_id="llm_dashboard_validation",
-            params={"dashboard_id": "dash-1", "threshold": 0.8},
-        )
-
-    def test_schedule_without_params(self):
-        """Schedule without extra params still works."""
-        mock_scheduler = MagicMock()
-        mock_task_manager = AsyncMock()
-
-        with patch("src.plugins.llm_analysis.scheduler.get_scheduler_service", return_value=mock_scheduler), \
-             patch("src.plugins.llm_analysis.scheduler.get_task_manager", return_value=mock_task_manager):
-
-            schedule_dashboard_validation(
-                dashboard_id="dash-2",
-                cron_expression="0 0 * * *",
-                params={},
-            )
-
-        job_func = mock_scheduler.add_job.call_args[0][0]
-        import asyncio
-        asyncio.run(job_func())
-
-        mock_task_manager.create_task.assert_called_once_with(
-            plugin_id="llm_dashboard_validation",
-            params={"dashboard_id": "dash-2"},
-        )
-# #endregion Test.LLMAnalysisScheduler

backend/tests/plugins/translate/test_scheduler.py

@@ -402,6 +402,9 @@ class TestExecuteScheduledTranslation:
             patch(
                 "src.plugins.translate.scheduler.NotificationService",
             ) as MockNotif,
+            patch(
+                "src.dependencies.get_async_job_runner",
+            ) as MockRunner,
         ):
             mock_orch = MagicMock()
             mock_run = MagicMock()
@@ -412,7 +415,12 @@ class TestExecuteScheduledTranslation:
             mock_orch.execute_run.side_effect = RuntimeError("LLM call failed")
             MockOrch.return_value = mock_orch
 
-            # Provider mock — send must return a real coroutine for asyncio.run()
+            # Mock the runner to raise the exception from execute_run
+            mock_runner = MagicMock()
+            mock_runner.run.side_effect = RuntimeError("LLM call failed")
+            MockRunner.return_value = mock_runner
+
+            # Provider mock — send must return a real coroutine for runner.run()
             async def _dummy_send(*args, **kwargs):
                 pass
             mock_provider = MagicMock()