### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### The Feature

Prometheus metrics for the lifecycle of managed batches and file operations:

- Batch creation, completion, and deletion events - File upload size and operation duration - `CheckBatchCost` polling metrics (jobs polled, jobs processed, errors encountered per poll cycle) - Labels for model, provider, user, and status on each metric

### Motivation, pitch

We currently have **zero Prometheus visibility** into managed batch lifecycle or the `CheckBatchCost` background poller. We requested this back in February, and it has been attempted three separate times, each closed without merging:

- #22256 (closed unmerged) - #23797 (closed unmerged) - #24566 (closed unmerged Jul 18 — despite an initial "should get merged by today" from the vendor on Mar 25)

Without these metrics we have no way to alert on batch-poller failures, track batch throughput/latency, or catch a `CheckBatchCost` regression before a user reports missing spend data. Given batches are already one of our lowest-confidence workstreams (see the many other open batch bugs), lifecycle observa…