add fast loading methods for bigquery destination

[rudolfix]

Background
Current method using load jobs is optimized for large loads. However most of the load packages is quite small. We can improve loading speeds by implementing streaming insert copy jobs. There's a singer target https://github.com/z3z1ma/target-bigquery by @z3z1ma from where we can take code.

we'll implement unconstrained (schema-less) version of the above when #891 in merged. this is about extending existing destination

Tasks

1. • allow user to select the loading API via destination configuration and per resource/table via bigquery adapter
2. • implement insert api and optionally storage write api
3. • allow both parquet and jsonl to be loaded this way. consider adding standard file readers for those types (port them from data sink destination #891 )
4. • tests and documentation