feat(CF-2411): Pipeline hardening — Sentry, retry, concurrent limits, audit log

- Sentry transactions wrapping pipeline execution with tags
- Retry with exponential backoff for transient failures (connect, timeout, 5xx)
- Concurrent execution limit (3/user) enforced in scheduler
- Audit log events fired at each pipeline lifecycle point
- Resume support: skip already-completed steps on restart

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

cron/scheduler.py

@@ -13,6 +13,7 @@ logger = logging.getLogger(__name__)
 
 SYNC_INTERVAL = 300  # 5 minutes — full job reconciliation
 PENDING_CHECK_INTERVAL = 15  # 15 seconds — fast check for manual triggers
+MAX_CONCURRENT_PER_USER = 3  # CF-2411: prevent runaway pipelines
 
 
 class CronScheduler:
@@ -145,6 +146,16 @@ class CronScheduler:
         pipelines = await self._pipeline_state.fetch_active_pipelines()
         for pipeline in pipelines:
             if pipeline.get("lastStatus") == "pending":
+                # CF-2411: concurrent limit check
+                user_id = pipeline.get("userId", "")
+                if user_id:
+                    active = await self._pipeline_state.count_active_executions(user_id)
+                    if active >= MAX_CONCURRENT_PER_USER:
+                        logger.warning(
+                            "Pipeline %s skipped: user %s has %d active executions (limit %d)",
+                            pipeline["name"], user_id, active, MAX_CONCURRENT_PER_USER,
+                        )
+                        continue
                 logger.info("Pending pipeline trigger: %s", pipeline["name"])
                 asyncio.create_task(self.pipeline_engine.run(pipeline))
 
@@ -155,6 +166,16 @@ class CronScheduler:
                 sleep_secs = self._seconds_until_next_run(pipeline)
                 if sleep_secs > 0:
                     await asyncio.sleep(sleep_secs)
+                # CF-2411: concurrent limit check
+                user_id = pipeline.get("userId", "")
+                if user_id:
+                    active = await self._pipeline_state.count_active_executions(user_id)
+                    if active >= MAX_CONCURRENT_PER_USER:
+                        logger.warning(
+                            "Pipeline %s cron skipped: user %s at limit (%d/%d)",
+                            pipeline["name"], user_id, active, MAX_CONCURRENT_PER_USER,
+                        )
+                        continue
                 await self.pipeline_engine.run(pipeline)
         except asyncio.CancelledError:
             pass