kvserver: move test to ccl

[andrewbaptist]

Previously the test was validating the behavior with Global reads, however since the test was not in the ccl package it required complicated mangaling of the span configs. Now it simply uses the standard SQL commands to create the table.

Epic: none
Fixes: #119230

Release note: None

[cockroach-teamcity]

This change is 

[nvanbenschoten]

I'm not sure about this change. It doesn't feel like pkg/ccl/multiregionccl is the right home for a test that's primarily interested in the behavior of a function (maybeCommitWaitBeforeCommitTrigger) deep inside the pkg/kv/kvserver package.

Was the way we were configuring this the cause of the flakiness? Is there a better way that we can configure spanconfigs from within the kvserver package without passing through SQL?

Reviewed 1 of 3 files at r1, all commit messages.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @pav-kv)

[andrewbaptist]

I agree with your reluctance here, I didn't like this much either, and I wouldn't have put it there except that its not possible to create a table with global locality without it being in CCL.

The problem is due to the way that span configs today can only be manually set up through SQL and worse that we can only set the span configs on replicas. We could hold off on merging this, but it will require a fairly major change to the way we create span configs to do this correctly. I have some thoughts about how to do this, but the change is somewhat invasive. Specifically we would want to disable either the automatic span config creation or alternatively some type of manual SpanConfigSubscriber to get around this limitation.

I had also considered creating a ccl package within kvserver which was only for tests that required the ccl.

From a very long term perspective, we should draw a stronger line and not have KV depend on SQL at all, even for unit tests. That will require a lot of work to happen given where we are now. SpanConfigs particularly are at an ugly border.