Update Ci::PendingBuild async when shared runner is toggled

Ref: #346229 (closed)

What does this MR do and why?

Following up !75197 (merged), this MR fixes a cross database query between our 2 new databases identified as a ~"sharding-blocker" when a group is transferred.

Why are we doing this?

As explained on our multi databases documentation it will not be possible to make a transaction accessing our 2 databases.

This is why need to execute the Groups::UpdateSharedRunnersService outside of the Group.transaction.

To solve this problem we are leveraging .run_after_commit_or_now to ensure we first persist the group and close the transaction before the ci_pending_builds table which is will happen in separate database.

Here are the 2 places where our new callback will get triggered when updating the shared_runners setting for a group:

• UpdateService - happening outside of a transaction
• TransferService - happening in a transaction

Here a screenshot of this new behavior below 👇

MR acceptance checklist

This checklist encourages us to confirm any changes have been analyzed to reduce risks in quality, performance, reliability, security, and maintainability.

• I have evaluated the MR acceptance checklist for this MR.