feat(bigframes): Defer unnamed @udf deployment until needed

[TrevorBergeron]

Thank you for opening a Pull Request! Before submitting your PR, there are a few things you can do to make sure it goes smoothly:

• 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 the tests and linter pass
• Code coverage does not decrease (if any source code was changed)
• Appropriate docs were updated (if necessary)

Fixes #<issue_number_goes_here> 🦕

[gemini-code-assist]

Code Review

This pull request implements deferred deployment for unnamed User Defined Functions (UDFs) in BigFrames. Instead of provisioning UDFs immediately during registration, they are now represented as PythonUdf definitions and deployed only when the execution plan is prepared for BigQuery execution. The changes include adding tracking for deployed routines in the function session, a new plan-rewriting step in the caching executor to handle on-demand deployment, and updated data structures for UDF requirements. Feedback suggests parallelizing these deployments using asyncio.gather to improve performance when multiple UDFs are present in a single plan.

[gemini-code-assist]

UDFs are currently deployed sequentially. Since each deployment involves network calls to BigQuery and resource provisioning, this can significantly delay query execution when multiple UDFs are used in a single plan. Parallelizing these deployments using asyncio.gather would improve performance.

        # Deploy UDFs in parallel to improve performance
        tasks = [
            asyncio.to_thread(
                session._function_session.deploy_undeployed_udf,
                session,
                udf,
            )
            for udf in unique_undeployed_udfs
        ]
        results = await asyncio.gather(*tasks)
        deployed_mapping = dict(zip(unique_undeployed_udfs, results))