Patch CursorWrapper dynamically to allow multiple base classes.

[tim-schilling]

When the debug SQL logs are enabled, the wrapper class is CursorDebugWrapper not CursorWrapper. Since we have inspections based on that specific class they are removing the CursorDebugWrapper causing the SQL logs to not appear.

This attempts to dynamically patch the CursorWrapper or CursorDebugWrapper with the functionality we need.

This doesn't do a full regression test, but it may be possible to get it to work with:

TEST_ARGS='--debug-sql' make test

Which causes the current set of tests to fail since they are keyed to CursorWrapper.

Fixes #1814

Checklist:

• I have added the relevant tests for this change.
• I have added an item to the Pending section of docs/changes.rst.

[tim-schilling]

@living180 I can't add you to the requested reviewers, but I'd appreciate you taking a look at this PR.

[matthiask]

LGTM. Thanks!

[tim-schilling]

@matthiask sorry, can you review that last commit too. It's a bit of a hack, but it's either that or inspecting self._outcome.results to determine if the test is running with the debug sql option.

[tim-schilling]

I'm likely to cut a release after this PR.

[matthiask]

@matthiask sorry, can you review that last commit too. It's a bit of a hack, but it's either that or inspecting self._outcome.results to determine if the test is running with the debug sql option.

Thanks, I think the way it's implemented makes sense!

[living180]

Didn't have time to review this until just now, but the implementation looks good and a basic smoke test passes for me. The only improvement I'd suggest would be a bit of bike-shedding by renaming NormalCursorWrapper and ExceptionCursorWrapper and NormalCursorMixin and ExceptionCursorMixin, respectively, but that's just a matter of taste.

[tim-schilling]

Good enough for me.