Define BQ layout of clinical data tables

[fedorov]

Per discussion, we can define a new dataset, which would be versioned, and would contain:

• per-collection clinical data tables (e.g., bigquery-public-data.idc_clinical_v9.*); we agreed that it makes sense to keep prefix _clinical for the cases where there is a single clinical data table per collection, since there will be cases where there will be multiple such tables, and we would need to differentiate based on the suffix)
• "dictionary" defining data elements for all clinical data tables (1 table) (something that will be based on what currently is in idc-dev-etl.clinical4.clinical_meta)
• an equivalent of "auxiliary" table that will define per-table metadata attributes (such as source of the table, hash of the source file, IDC data version where it was introduced and modified) (we do not yet have an example or defined structure of this, see #29 )

[fedorov]

I think the above is how things are now implemented, and this issue can be closed.