feat(bigquery): support jobs.insert in createQueryArrowStream

[ohmayr]

Some query options only work with jobs.insert (for example jobId, labels, dryRun, or jobTimeoutMs), so createQueryArrowStream failed for those queries. This runs them through createQueryJob instead.

• Used when buildQueryRequest_ returns undefined (a jobs.insert-only option, or FAST_QUERY_PATH=DISABLED). timeoutMs and jobCreationMode only apply to jobs.query, so they stay out of the job configuration; timeoutMs bounds the wait instead.
• Sets stream.jobReference and emits the job event.
• A dryRun emits one 0-row, 0-column RecordBatch without waiting or calling ReadRows; the job's metadata holds the dry-run statistics.
• Otherwise waits for the job to finish and streams from streams/_default at offset = 0. Errors thrown synchronously by createQueryJob destroy the stream.

[gemini-code-assist]

Code Review

This pull request implements a fallback to route queries that cannot be executed via jobs.query through jobs.insert (createQueryJob) in createQueryArrowStream_, and adds comprehensive unit and system tests to verify this behavior. Feedback was provided to ensure type safety by guarding against an undefined job object before calling waitForJobCompletion and to use the idiomatic stream.push(null) instead of stream.end() to signal EOF.

[quirogas]

If timeoutMs: 0 is passed, Job#getQueryResults (src/job.ts:627) sends timeoutMs=0 (returning jobComplete: false immediately in 0ms) and does not error because if (timeoutOverride) is falsy for 0, causing poll() to loop with zero delay. Only forward timeoutMs when it's truthy (> 0):

Suggested change

	{
	autoPaginate: false,
	maxResults: 0,
	...(timeoutMs !== undefined && {timeoutMs}),
	},
	{
	autoPaginate: false,
	maxResults: 0,
	...(timeoutMs ? {timeoutMs} : {}),
	},

[ohmayr]

Done, timeoutMs is only forwarded when it's non-zero.

[quirogas]

Strip jobCreationMode alongside timeoutMs (as noted in the comment above), and wrap bq.createQueryJob in a try / catch routing synchronous errors (such as invalid params or destination) to stream.destroy(err as Error):

Suggested change

	// clustering). jobs.insert does not accept queryResultsFormat or
	// jobCreationMode.
	// timeoutMs only bounds the wait below, so keep it out of the job
	// configuration sent to jobs.insert.
	const {timeoutMs, ...jobQuery} = queryObj;
	bq.createQueryJob(jobQuery, (err, job, resp) => {
	// clustering). jobs.insert does not accept queryResultsFormat or
	// jobCreationMode.
	// timeoutMs only bounds the wait below, so keep it out of the job
	// configuration sent to jobs.insert.
	const {timeoutMs, jobCreationMode: _jobCreationMode, ...jobQuery} =
	queryObj;
	bq.createQueryJob(jobQuery, (err, job, resp) => {

[ohmayr]

Wrapped createQueryJob in a try/catch. I left jobCreationMode in because jobs.insert accepts and drops it (checked with a live createQueryJob({jobCreationMode: 'JOB_CREATION_OPTIONAL', ...}); it isn't in the returned configuration.query), and BigQuery#query passes it through the same way.

[ohmayr]

Update: I dropped jobCreationMode from the jobs.insert configuration after all, to match Python, which only sends it on jobs.query. BigQuery ignores it on jobs.insert, so nothing changes for callers; the unit test now sets it and checks it's left out.