[Service] Implement BigQuery (REST + Storage gRPC)

[hectorvent]

Summary

Implement BigQuery — the GCP analytics warehouse. Highest-value data service; community emulators (e.g. goccy/bigquery-emulator) prove demand. The main cost is a SQL engine, so this should be phased.

Protocol / transport

• Core API: REST (Discovery bigquery.googleapis.com) — datasets, tables, jobs, tabledata
• Storage Read/Write API: gRPC google.cloud.bigquery.storage.v1

Endpoint / auth bypass

BIGQUERY_EMULATOR_HOST (supported by some clients) / client apiEndpoint override.

Authoritative reference

• REST Discovery document for bigquery.googleapis.com
• Proto: local/google/googleapis/google/cloud/bigquery/storage/v1/storage.proto
• SDK: local/google/google-cloud-java/java-bigquery, java-bigquerystorage

Scope (phased)

• Phase 1 (metadata + ingest): Datasets CRUD, Tables CRUD + schema, tabledata.insertAll (streaming inserts), trivial SELECT * / COUNT.
• Phase 2 (SQL): embed a GoogleSQL-compatible engine for jobs.query / jobs.insert(QUERY) + getQueryResults. Document dialect deviations explicitly.
• Phase 3: Storage Write API (gRPC).

Parity notes

• Job lifecycle (PENDING/RUNNING/DONE) + getQueryResults paging.
• GoogleSQL dialect; surface unsupported SQL as a clear error rather than silent divergence.

Test plan

java-bigquery: create dataset/table, insertAll, run a query and read results back; assert schema + row values.

---

Implementation guidance: Follow the google-parity workflow — read the authoritative proto/SDK under local/google/* before implementing, mirror the wire shape exactly, and validate with GCP SDK-based compatibility tests (compatibility-tests/sdk-test-java). See AGENTS.md → "Adding a New GCP Service". Keep storage namespaced by project ID via StorageFactory.

Upstream references (mirrored locally under local/google/*):

• googleapis (proto definitions): https://github.com/googleapis/googleapis
• google-cloud-java (Java SDK source): https://github.com/googleapis/google-cloud-java