BUG: potential resource leak when using BigQuery Storage API

[tswast]

It came to my attention in googleapis/google-cloud-python#9457 that when a gRPC Google Cloud client object is destroyed, it doesn't clean up all the underlying resources from the gRPC transport layer.

I'd like to address this at the Google Cloud client level by providing better support for short-lived gRPC clients, but in the meantime, we need to call bqstorage_client.transport.close() when we are done using a BigQuery Storage API client.