feat(source/bigquery): attach SQLCommenter attributes as BigQuery job labels

[spark2ignite]

Description

Enabling --sql-commenter currently has no effect on the BigQuery source:
the commenter is only wired into postgres/mysql-family sources, and BigQuery
would not retain SQL-text comments in job metadata anyway. As noted on the
issue, today exactly one attribute reaches the job — the automatic
mcp-toolbox-tool label added in v1.5.0 (#1975), which carries the tool
type (e.g. bigquery-execute-sql). The seven remaining SQLCommenter
attributes — per-user, per-agent, per-model attribution and the trace
linkage — never reach the job. This PR adds them. The SQLCommenter
tool.name attribute (the MCP tool name, e.g. execute_sql) is distinct
from the existing type label; both are kept, and explicit tool-supplied
labels always win on key collisions.

Following the direction discussed in the linked issue, this PR attaches the
SQLCommenter attributes as native job labels instead of prepending a
comment, so they surface first-class in INFORMATION_SCHEMA.JOBS, audit
logs, and billing exports — no query text parsing needed to recover them.

• New sqlcommenter.Labels(...) returns the exact attribute set
  PrependComment uses (tool.name, client, client.model,
  client.user.id, client.agent.id, traceparent, server,
  db.system.name), sanitized to BigQuery label constraints: lowercase
  [a-z0-9_-], max 63 chars, keys start with a letter, dots mapped to
  underscores (tool.name → tool_name).
• Labels are merged in Source.RunSQL, so both bigquery-execute-sql and
  bigquery-sql are covered with no per-tool changes; ML/AI tools that don't
  flow through RunSQL are unaffected, matching the scope note on the issue.
  Explicit tool-supplied labels (e.g. mcp-toolbox-tool) win on collisions.
• Gated identically to other sources: global --sql-commenter flag plus an
  optional per-source sqlCommenter override in tools.yaml, default off.
• Docs: new "BigQuery: Job Labels" section in the SQL Commenter monitoring
  page (with an example INFORMATION_SCHEMA.JOBS attribution query) and a
  sqlCommenter row in the BigQuery source reference.

Tests: table-driven unit tests for key/value sanitization (uppercase,
illegal characters, truncation, leading non-letter keys), enable/override
precedence mirroring the existing PrependComment cases, label merge
precedence, and YAML config parsing. go test -race ./cmd/... ./internal/...
and golangci-lint run pass.

Also verified end-to-end against a live BigQuery project: with
sqlCommenter: true, a query issued through the MCP protocol (JSON-RPC
tools/call with _meta["dev.mcp-toolbox/telemetry"] set) produced a job
carrying all eight attributes as labels alongside the existing
mcp-toolbox-tool:

"labels": {
  "client": "select-e2e-client_0_0_1",
  "client_agent_id": "agent-e2e-42",
  "client_model": "test-model-1",
  "client_user_id": "vadim-e2e",
  "db_system_name": "bigquery",
  "mcp-toolbox-tool": "bigquery-execute-sql",
  "server": "genai-toolbox_1_9_0_dev_darwin_arm64",
  "tool_name": "execute_sql",
  "traceparent": "00-f58710a2aea7bf757cf9292b026a2bce-de6a2f6214d3f085-01"
}

With the flag off, the job carries only mcp-toolbox-tool — byte-identical
to current behavior. The labels are queryable in
INFORMATION_SCHEMA.JOBS_BY_PROJECT (grouping bytes billed by tool_name
and client works as expected).

PR Checklist

• Make sure you reviewed
  CONTRIBUTING.md
• Make sure to open an issue as a
  bug/issue
  before writing your code! That way we can discuss the change, evaluate
  designs, and agree on the general idea
• Ensure you have manually reviewed the entire diff before requesting a
  review
• Ensure the tests and linter pass
• Code coverage does not decrease (if any source code was changed)
• Appropriate docs were updated (if necessary)
• Make sure to add ! if this involve a breaking change

🛠️ Fixes #3286

[gemini-code-assist]

Code Review

This pull request introduces SQL Commenter support for the BigQuery source. Because BigQuery does not retain SQL-text comments in its logs, SQL Commenter attributes are attached as native job labels instead. The implementation includes sanitization logic to satisfy BigQuery's label constraints, a helper to merge explicit and commenter-derived labels, configuration updates, and comprehensive unit tests and documentation. No review comments were provided, so there is no additional feedback.

[spark2ignite]

/gemini review

[gemini-code-assist]

Code Review

This pull request adds support for SQL Commenter in BigQuery by attaching telemetry attributes as native job labels instead of SQL-text comments, updating the configuration, documentation, and adding sanitization logic for BigQuery label constraints. The review feedback highlights a potential issue where sanitized label values starting with an underscore or dash would be rejected by BigQuery, and suggests prefixing them with 'x' along with adding corresponding test cases.