support clustering on BigQuery

[drewbanin]

Feature

Feature description

BigQuery recently announced support for table clustering. With this new feature, a list of up to four "clustering columns" can be supplied to a create table as (...) statement. BigQuery will use the clustering columns to colocate similar data. This should speed up filters and aggregations, and BigQuery can both skip scanning whole blocks of data, as well as avoid shuffling data between nodes for certain aggregations. Docs: https://cloud.google.com/bigquery/docs/creating-clustered-tables

Who will this benefit?

This feature will benefit users of BigQuery. It will work in concert with table partitioning.

Proposed Implementation

This feature can be implemented in the BigQuery implementation of create_table_as(). It's implementation will look similar to the partition_by implementation.

[beckjake]

cluster_by can't be used without partition_by. Should we enforce that in the macro, or let the database return the error when a user doesn't specify both?

[drewbanin]

Yeah, I'm fine with letting BQ return an error in that scenario

[beckjake]

Fixed in #978