fix: web console services card missing broker count without SQL

[wlswo]

Description

The home dashboard's Services card has two code paths. The SQL path queries sys.servers (discovery-backed, so every
node role is reported), while the coordinator-only fallback only counted historical, middle_manager, and peon.
/druid/coordinator/v1/servers?simple does return brokers as well (brokers are tracked by
InventoryView/ServerType.BROKER), but they were dropped before being passed to the card, so non-SQL deployments
showed no broker count even when brokers were present.

This change adds the broker count to the fallback so its output matches the SQL path.

The other roles missing from the fallback (coordinator, overlord, router, indexer, middle_manager vs
indexer-executor distinction) cannot be derived from /druid/coordinator/v1/servers?simple and would require
additional endpoints; they are intentionally left out of this change.

Release note

The home view's Services card now correctly shows the broker count on clusters where the web console is talking to
the coordinator without SQL access.

Key changed/added classes in this PR

• web-console/src/views/home-view/services-card/services-card.tsx

---

This PR has:

• been self-reviewed.
• added documentation for new or modified features or behaviors. (N/A — UI count fix, no user-facing docs)
• a release note entry in the PR description.
• added Javadocs for most classes and all non-trivial methods. (N/A — TypeScript)
• added or updated version, license, or notice information in
  licenses.yaml (N/A — no dependency change)
• added comments explaining the "why" and the intent of the code wherever would not be obvious for an unfamiliar
  reader. (N/A — one-line change)
• added unit tests or modified existing tests to cover new code paths, ensuring the threshold for code
  coverage is met. (no new test —
  existing snapshot test covers the SQL path; the fallback path is a one-line filter mirroring an existing line)
• added integration tests. (N/A)
• been tested in a test Druid cluster. (local npm run typecheck / npm run jest / npm run eslint / npm run prettify-check all pass)

[wlswo]

Closing in favor of #19481 after verifying this against a live Druid 30 cluster.

/druid/coordinator/v1/servers?simple is backed by InventoryView and only reports segment-loading servers —
brokers don't appear in it unless they hold broadcast segments, which is rare in practice. Confirmed empirically: on
the cluster I tested, the endpoint returned only historical and indexer-executor entries even though the cluster
had a running broker on a separate host.

So the broker filter added in this PR would always count zero. The real bug is broader: the coordinator-only fallback
was using the wrong endpoint, and overlord, coordinator, router, broker, and indexer were all silently
dropped — not just broker. The new PR switches the endpoint to /druid/coordinator/v1/cluster (discovery-backed,
same source sys.servers uses) and brings the fallback in line with the SQL branch.

Sorry for the noise.