Convert select query extenders to backend-overrideable services

Unpublished CR - this was not done when the commit was reverted.

mondrake → closed merge request !4408

MR 4452 is another attempt at this. Starting from the previous approach, this uses service aliasing to allow keeping calling the extension by class name, and finally deprecates Connection::getDriverClass().

The solution looks good to me.

I think I addressed all of @daffie's points. Thanks.

The last couple of nitpicks.
After this it is RTBC for me.

mondrake → closed merge request !4409

All code changes look good to me.
All my points have been addressed.
The IS and the CR are in order.
For me it is RTBC.

I'm triaging RTBC issues → . I workd on this a few days ago as well.

I corrected a typo in the release snippet I saw when reading the Issue Summary.

I read the comments the other day and refreshed my memory just now. I did not find any unanswered queries there. The only thing is the followup to create at some point.

I am leaving this at RTBC.

The Needs Review Queue Bot → tested this issue. It fails the Drupal core commit checks. Therefore, this issue status is now "Needs work".

This does not mean that the patch needs to be re-rolled or the MR rebased. Read the Issue Summary, the issue tags and the latest discussion here to determine what needs to be done.

Consult the Drupal Contributor Guide → to find step-by-step guides for working with issues.

Rebased and added @phpstan-ignore-next-line to calls of deprecated methods that are left in for BC purposes. Back to RTBC.

Rebased after commit of the transaction refactor issue.

Updated the branch/version on the CR. I did not review the CR.

The followup is still needed. It's a followup, not a blocker, so we can still go ahead here. We've also made coding standards progress on allowing the protected property promotion and the multi-line constructor format.

We do still need that additional framework manager feedback.

Meanwhile, there are various quite small coding standards issues with the MR, plus any number of inline comments we don't actually need. Plus, the scope of the D11 removal of the BC layer is sufficiently beyond a simple deprecation that we should go ahead and file the D11 followup now. (An existing issue was closed as outdated -- either it should be reopened and rescoped to be a part of the D11 beta requirements, or it should be replaced by an issue that is.)