Add grouped nodeTypeCounts query; use COUNT(*) for totalCount

[shangyian]

Summary

The namespace landing page shows per-type filter pills ("Metric (342)", "Source (…)" etc). Today the UI populates them by firing one findNodesPaginated(nodeTypes:[X], limit:1){ totalCount } per node type, which results in 5 separate count scans per page load (aliased into a single request, but still N scans server-side).

This PR adds a generic server-side primitive to do it in one grouped scan, plus a small fix to the count query that backs every paginated totalCount.

New generic nodeCounts query

nodeCounts(groupBy: NodeGroupBy!, namespace: String): [NodeCount!]!

type NodeCount { value: String!, count: Int! }
enum NodeGroupBy { TYPE }

This runs the existing filter builder with GROUP BY <column> and returns {value: count} in one query. We deliberately kept this as a separate field, not folded into findNodesPaginated since counting and pagination are different concerns. NodeGroupBy starts at TYPE, but adding STATUS/MODE later is just an enum value + a column mapping.

COUNT(*) instead of COUNT(DISTINCT id) for totalCount

Node.count_by (which backs every paginated totalCount) used COUNT(DISTINCT Node.id). However, COUNT(*) is exactly equivalent and skips the sort/hash DISTINCT forces. This change yields identical results, and it stops degrading with table size.

Test Plan

• PR has an associated issue: #
• make check passes
• make test shows 100% unit test coverage

Deployment Plan

[netlify]

✅ Deploy Preview for thriving-cassata-78ae72 canceled.

Name	Link
🔨 Latest commit	de802af
🔍 Latest deploy log	https://app.netlify.com/projects/thriving-cassata-78ae72/deploys/6a4a81402a097600098d03f8