[PULP-2474] Optimize Maven repository metrics counts

[kdelee]

The repository metrics endpoint runs three separate COUNT(DISTINCT) queries over MavenPackage content. Combine them into one Django aggregate query using PostgreSQL row-valued DISTINCT expressions.

This preserves the API and exact results while reducing production-scale GABI latency from about 9.4–9.8s to 3.6–3.8s under three concurrent requests.

Tests:

• Ruff format/check
• 234 unit tests passed, 7 skipped
• 2 focused functional tests passed in the OCI container

[kdelee]

GABI benchmark compared the existing three-query implementation with the new single-query Django aggregate.

Existing SQL pattern:

  SELECT COUNT(*)
  FROM (
    SELECT DISTINCT group_id, artifact_id
    FROM maven_mavenpackage
    WHERE content_ptr_id IN (
      SELECT unnest(content_ids)
      FROM core_repositoryversion
      WHERE pulp_id = '<version>'
    )
  ) packages;

  SELECT COUNT(*)
  FROM (
    SELECT DISTINCT
      group_id,
      artifact_id,
      REGEXP_REPLACE(version, '-[^-]+$', '')
    FROM maven_mavenpackage
    WHERE content_ptr_id IN (
      SELECT unnest(content_ids)
      FROM core_repositoryversion
      WHERE pulp_id = '<version>'
    )
  ) versions;

  SELECT COUNT(*)
  FROM (
    SELECT DISTINCT group_id, artifact_id, version
    FROM maven_mavenpackage
    WHERE content_ptr_id IN (
      SELECT unnest(content_ids)
      FROM core_repositoryversion
      WHERE pulp_id = '<version>'
    )
  ) builds;

The API executed these as three separate database queries.

New SQL shape:

  SELECT
    COUNT(DISTINCT (group_id, artifact_id)),
    COUNT(DISTINCT (
      group_id,
      artifact_id,
      REGEXP_REPLACE(version, '-[^-]+$', '')
    )),
    COUNT(DISTINCT (group_id, artifact_id, version))
  FROM maven_mavenpackage
  WHERE content_ptr_id IN (
    SELECT unnest(content_ids)
    FROM core_repositoryversion
    WHERE pulp_id = '<version>'
  );

Results matched exactly.

Single-request results:

  Existing implementation:
    cold: ~4.86s for first query
    warm: ~3.0s per query
    ~9s total across all three queries

  New implementation:
    cold: 4.86s
    warm: 3.41s, 3.39s

Three concurrent requests:

  Existing implementation: 9.4–9.8s per request
  New implementation:       3.6–3.8s per request

Summary:

  Same counts
  One DB query instead of three
  Approximately 60% faster under concurrent polling
  Less database contention
  No API response or schema change

[patchback]

Backport to 0.34: 💚 backport PR created

✅ Backport PR branch: patchback/backports/0.34/cdfe60dee5cc2a112646d495b7660796dbc7fc6d/pr-541

Backported as #542

🤖 @patchback
I'm built with octomachinery and
my source is open — https://github.com/sanitizers/patchback-github-app.