SQL Targets: Inserts fail if activate_version message has added _sdc columns

Summary

The INSERT statement generated by SQLSink.bulk_insert_records does not specify target column names. This is fine when the incoming records contain all the columns in the table but fails when other columns exist in the table - for example the _sdc* columns created by the processing of an activate_version record. While I haven't tested it explicitly this would also be a problem if the stream's schema changed over time, e.g. a column was removed from the source table.

Steps to reproduce

Run the unit test added for !287 (merged) ;)

What is the current bug behavior?

The INSERT fails because values are not specified for all columns.

What is the expected correct behavior?

The INSERT should succeed and populate the columns found in the incoming stream.

Possible fixes

MR incoming