Track merged_commit_sha on merge

What does this MR do and why?

This ties together a couple of recent changes to ensure fast-forward merge trains merges use the correct SHAs on merge, and that the associated merge request gets updated with the correct values:

In !129311 (merged) we added the column merged_commit_sha, which represents the SHA of every merged merge request, independently of the merge method.

In !129979 (merged), we added train_ref to merge_requests.merge_params, which contains the exact SHAs used to construct the ref, which should be written to the merge request on merge.

This change is behind the existing feature flag fast_forward_merge_trains_support, see #28244

Addresses both #422483 (closed) and #418822 (closed)

Screenshots or screen recordings

Before	After
Without this change, the merge train gets re-created on each non-squashed fast-forward merge	After this change, the merge train does not get re-created on each merge (each passed pipeline implies a merge. You can verify this too by checking the commits. The seeded project has no pipeline on main)

How to set up and validate locally

1. Enable the following feature flags:

   Feature.enable(:fast_forward_merge_trains_support)
   Feature.enable(:merge_trains_create_ref_service)

2. Seed a project with a bunch of merge requests:

   bin/rake gitlab:seed:merge_trains:project

3. Visit the project, go to Settings > Merge Requests and change the Merge method to Fast-forward.
4. Visit a few of the open merge requests in separate tabs or windows
5. Wait for the pipeline to finish on all of them (wait for the button Set to auto-merge to change to Merge)
6. In rapid succession, press Merge for each of them to start a Fast-forward merge train
7. Navigate to Build > Pipelines in the sidebar
8. Verify that none of the started merge trains has a merge commit
9. Verify that all of the merge trains get merged and that the train does not get re-created at any point

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.

For database review

Example raw query after the change, see execution on postgres.ai (internal link):

SELECT
    1 AS one
FROM
    "merge_requests"
WHERE
    "merge_requests"."id" IN (
        SELECT
            "merge_trains"."merge_request_id"
        FROM
            "merge_trains"
        WHERE
            "merge_trains"."target_project_id" = 7764
            AND "merge_trains"."target_branch" = 'master'
            AND ("merge_trains"."status" IN (1, 4))
        ORDER BY
            "merge_trains"."id" DESC
        LIMIT 20)
AND (merge_commit_sha = '123'
    OR in_progress_merge_commit_sha = '123'
    OR squash_commit_sha = '123'
    OR merged_commit_sha = '123')
LIMIT 1