fix(google-cloud): a BigQuery table created by a concurrent run is not an error

[aaaaahaaaaa]

Summary

BigQueryDestination._create_table raised on a 409 Conflict from BigQuery. Two runs writing
different partitions of the same asset whose table does not exist yet (a backfill, or a queue burst
after downtime, or two runs queued back to back on a fresh destination) both see no table, both create
it, and the loser fails with Already Exists, which fail-fast turns into a failed run with every
other asset canceled. _ensure_dataset already tolerated the same race for datasets; _create_table
now does the same and loads into the table that just appeared, which has the schema it wanted.

Found while materializing a fresh destination locally: two job runs for consecutive days executed in
parallel and the second failed on facebook_ads.campaigns__<account>. Production schedules one run
per job per day, which is why it has not shown there.

Verification

• New test: get_table reports no table, create_table raises Conflict, the rows still load.
• uv run --frozen pytest packages/interloper-google-cloud: 124 passed. ruff and ty clean.

By Digitl

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!