Add proto specs for usage alerts

[bduffany]

Adds protos for basic usage alerting. For v0, I'm thinking the system can look something close to this:

• Users can manage* alerts using an "Alerting" tab in the Usage page, setting SKU + optional labels, alerting window (daily, weekly, monthly), and alerting thresholds after which the alerts will fire.
• Alerts can be stored in the primary DB (mysql).
• The rule evaluator can run as a k8s cronjob, running at some reasonable, short-ish interval. At each interval, the evaluator:
  • (1) queries UsageAlertingRules that need to be evaluated (given the current timestamp, last_fired_at timestamp, and alerting window)
  • (2) queries the RawUsage table to compare counts vs thresholds
  • (3) sends email alerts** if any thresholds are exceeded
  • (4) updates the last_evaluated_at and last_fired_at timestamps.

* Only adding Create, List, and Delete APIs for now. "Update" can be done by deleting and recreating a rule (could maybe add a "clone" button in the UI to make it slightly less tedious, or implement Update later if we have a need)
** If the cronjob fails between steps 3 and 4, we may send duplicate alerts (at-least-once semantics). As a "nice to have," if we choose an email delivery provider that supports idempotency keys, then we can get exactly-once delivery semantics.

[Copilot]

Pull request overview

This PR extends the usage protobuf API surface to support usage-based alerting rules, including rule metadata, configuration, status, and basic CRUD request/response messages.

Changes:

• Added new usage alerting rule protos (UsageAlertingRule, metadata/config/status, and UsageAlertingWindow enum) to proto/usage.proto.
• Added request/response messages for listing, creating, and deleting usage alerting rules.
• Updated Bazel proto build targets to include new proto dependencies (user ID + timestamp) for usage proto generation.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
proto/usage.proto	Adds message/enum definitions and RPC message shapes for usage alerting rules, including timestamps and creator identity.
proto/BUILD	Wires in user_id and timestamp deps for proto / TS / Go generation of the updated usage protos.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[maggie-lou]

Another type of alert that may be popular is "Alert if this metric is X% higher than last week". Even if we don't support that in v0, the proto should be flexible enough to add that later

In check-metrics.yaml, we have a concept of "absolute" and "relative" thresholds. Maybe something similar could apply here?

[bduffany]

Renamed to "absolute_threshold" and we can add "relative_threshold" later 👍

[maggie-lou]

Discussed this a bit in our meeting, but could we use a scheduled Workflow in the -internal repo rather than a k8s cron job? For example, we'd create one scheduled workflow per alert. Then we wouldn't have to re-implement some of the logic around when to evaluate the rules

This should also make editing the rules a bit easier. It does not sound like a pleasant UI experience to clone and delete rules to edit them 😅

[bduffany]

open to it - once scheduled workflows are battle-tested in prod, we could consider switching this over fully. In the meantime we could maybe extract a cronutil package once we see a common pattern emerging in the evaluation logic