Expand Firebase Firestore Agent Skill evaluation suite with multi-domain benchmarks.

[copybara-service]

Expand Firebase Firestore Agent Skill evaluation suite with multi-domain benchmarks.

Expands the Firebase Firestore Agent Skill evaluation suite with 5 specialized Web SDK evaluation benchmark suites (43 evaluation cases):

• classic_sdk: Core CRUD, document references, batch writes, queries, composite index configuration, cursor pagination, transactions, and aggregate queries.
• edition_routing: Standard vs Enterprise database detection, CLI provisioning, and feature gating.
• pipeline_api: Server-side relational joins, scalar expressions, vector similarity search, full-text search, and pipeline DML mutations.
• security_rules: Type checking, schema constraints, role-based access control, and helper function validation.
• fixtures & validators: Hermetic package setup, TypeScript compiler type-checking (tsc --noEmit), and mock Firebase CLI inspection.

Benchmark Results:

• Gemini 3.5 Flash 5-Suite Ablation: 97.6% candidate (41/42) vs 81.0% baseline (34/42), achieving +16.7% net lift and +20.6% relative lift.
  Report: https://evalin.corp.google.com/report/groupagent-firestore-clients-dev/25408bbcc2934582
• Production Gemini Flash 5-Suite Ablation: 90.5% candidate (38/42) vs 76.2% baseline (32/42), achieving +14.3% net lift.
  Report: https://evalin.corp.google.com/report/groupagent-firestore-clients-dev/5d0fe5a8a2a942d3

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.