fix: fix broker segment metadata cache refresh

[jtuglu1]

Description

Brokers can maintain a schema cache via segment metadata queries. Currently, if any of these queries timeout, the remaining queries are aborted until the next refresh. If you have a huge datasource delta (think 500k+ segments being scanned), such a query can fail/timeout and cause other unrelated datasources' broker schema discovery to fail. Without centralized schema through coordinator, there is no intra-datasource atomicity guarantee w.r.t schema discovery (it is just ASAP), so decoupling this error dependency and instead emiting a metric per datasource when failures occur.

Introduces segment/schemaCache/refresh/failed metric with a dataSource dimension, emitted when a refresh fails. Can alternatively just aggregate and emit at the end. Also open to keeping this a warning/error log.

Release note

Fix broker segment metadata cache refresh

---

This PR has:

• been self-reviewed.
  • using the concurrency checklist (Remove this item if the PR doesn't have any relation to concurrency.)
• added documentation for new or modified features or behaviors.
• a release note entry in the PR description.
• added Javadocs for most classes and all non-trivial methods. Linked related entities via Javadoc links.
• added or updated version, license, or notice information in licenses.yaml
• added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar reader.
• added unit tests or modified existing tests to cover new code paths, ensuring the threshold for code coverage is met.
• added integration tests.
• been tested in a test Druid cluster.

[FrankChen021]

Severity	Findings
P0	0
P1	0
P2	1
P3	0
Total	1

Severity	Findings
P0	0
P1	0
P2	1
P3	0
Total	1

Reviewed 4 of 4 changed files.

---

This is an automated review by Codex GPT-5.5

[FrankChen021]

[P2] Do not treat every QueryInterruptedException as shutdown

This catch rethrows all QueryInterruptedException instances before the per-datasource failure handler. In Druid, QueryInterruptedException is also used as a generic client-side wrapper for failed queries, including remote QueryExceptions with unknown/runtime failures, not only actual thread interruption. When one datasource's SegmentMetadataQuery fails through that wrapper, refreshSegments still aborts the whole cycle, later datasources are skipped, and no segment/schemaCache/refresh/failed metric is emitted for the offending datasource. Please distinguish real local interruption/cancellation from ordinary query failure, for example by checking the cause/errorCode, and let non-interruption query failures follow the isolated datasource path.