When getting a "Cannot query over table (...) without a filter over
column(s) (...)", the re-raising happening gobbles up the error message
and the context is lost.

This may or may not be the right fix, but clearly highlights the issue
of `self._query_job.errors` being `None` in that particular context.
When that is the case, the only context available is that of
`DatabaseError` without any details which clearly isn't enough detail.

It seems like wrapping the exception into another exception works as
expected and shows both stack traces as shown here:

The errors on the job resource are mostly redundant with the exception information.