Use Storage write API in BigQuery connector

[ebyhr]

Release notes

(x) Release notes are required, with the following suggested text:

# BigQuery
* Improve performance when writing rows. ({issue}`18897`)

[anusudarsan]

should the data be batched based on a request size limit (config)? https://cloud.google.com/bigquery/quotas#write-api-limits

[hashhar]

good to do but pre-existing issue.

[hashhar]

we are creating a new stream per Page.
BigQuery has a limit of 1k streams open at a time and 30k/4 hours (7.5k creations per hour).

Let's either document this if we think it's shouldn't be a problem or let's consider creating a single stream per page-sink/tablewriter task.

Also do we consider to use "pending mode" long term to provide proper isolation? With current mode ("committed") if a single stream fails then writes from other streams will still succeed and be visible. i.e. it's not ACID and there's no way to rollback. Note that this is same behaviour that we already had so it's not a regression in that sense.

[hashhar]

looks good to me

[hashhar]

nit: multi-line

Suggested change

	CreateWriteStreamRequest createWriteStreamRequest = CreateWriteStreamRequest.newBuilder().setParent(tableName.toString()).setWriteStream(stream).build();
	CreateWriteStreamRequest createWriteStreamRequest = CreateWriteStreamRequest.newBuilder()
	.setParent(tableName.toString())
	.setWriteStream(stream)
	.build();

[hashhar]

@ebyhr Does the docs need updating about any new IAM permissions which are needed.