feat: Support BQ decimal precision and scale for schema validation

[nehanene15]

• Limit Teradata query to one row for custom query schema generation (Avoids full table scan)
• Support non-defualt BigQuery precision/scale (Closes Schema validation: BigQuery Decimal Precision #883 and Closes Schema Validation : Bigquery Numeric #882 )

[nehanene15]

/gcbrun

[nj1973]

LGTM aside from comment about Hive (and maybe other engines) having the full scan issue fixed here for Teradata.

[nj1973]

We might also have this problem for other engines, e.g: Hive:

def _get_schema_using_query(self, query):
    # Removing LIMIT 0 around query since it returns no results in Hive
    cur = self.raw_sql(query)
    cur.fetchall()
    ibis_fields = self._adapt_types(cur.description)
    cur.release()

    return sch.Schema(ibis_fields)

We should be able to use fetchone in place of fetchall for this. Or embed in a subquery. I didn't check all engines so Hive may not be the only one.

Happy for this to be spun into another issue if you need to move quickly on this one.

[nehanene15]

Good catch! I'll update this for Hive and add a LIMIT 1 similar to the Ibis impala connection. Spanner only materializes one row, and all other DBs are handled by Ibis.