Fix dlx checkout command redirection (backport #16203)

[mergify]

This commit supersedes #15548.

What?

Fix the following genuine CI flake:

make -C deps/rabbit ct-rabbit_fifo_dlx_integration t=cluster_size_3:single_dlx_worker

Sometimes this test case failed with the logs showing the following:

2026-02-24 09:06:19.413770+00:00 [debug] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': vote granted for term 3 votes 2
2026-02-24 09:06:19.414048+00:00 [debug] <0.2377.0> started rabbit_fifo_dlx_worker <0.2600.0> for queue 'single_dlx_worker_source' in vhost '/'
2026-02-24 09:06:19.414096+00:00 [notice] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': candidate -> leader in term: 3 machine version: 7, last applied 5
2026-02-24 09:06:19.414388+00:00 [debug] <0.2602.0> queue 'single_dlx_worker_source' in vhost '/': updating leader record to current node rmq-ct-cluster_size_3-1-28000@localhost
2026-02-24 09:06:19.414279+00:00 [info] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': leader saw request_vote_rpc from {'%2F_single�[118;1:3u_dlx_worker_source','rmq-ct-cluster_size_3-3-28144@localhost'} for term 4 abdicates term: 3!
2026-02-24 09:06:19.417479+00:00 [notice] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': leader -> follower in term: 4 machine version: 7, last applied 5
2026-02-24 09:06:19.417533+00:00 [debug] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': is not new, setting election timeout.
2026-02-24 09:06:19.417740+00:00 [info] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': declining vote for {'%2F_single_dlx_worker_source','rmq-ct-cluster_size_3-3-28144@localhost'} for term 4, candidate last log {index, term} was: {5,2}  last log entry {index, term} is: {{6,3}}
2026-02-24 09:06:19.417824+00:00 [debug] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': leader call - leader not known. Command will be forwarded once leader is known.
2026-02-24 09:06:19.418190+00:00 [debug] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/' declining pre-vote to {'%2F_single_dlx_worker_source','rmq-ct-cluster_size_3-2-28072@localhost'} for term 3, current term 4
2026-02-24 09:06:19.428043+00:00 [debug] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': resetting last index to 5 from 6 in term 4
2026-02-24 09:06:19.428157+00:00 [info] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': detected a new leader {'%2F_single_dlx_worker_source','rmq-ct-cluster_size_3-3-28144@localhost'} in term 4
2026-02-24 09:06:19.428280+00:00 [debug] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': mem table overwriting detected whilst staging entries, opening new mem table
2026-02-24 09:06:19.436299+00:00 [debug] <0.2377.0> queue 'single_dlx_worker_source' in vhost '/': enabling ra cluster changes in 4, index 6
2026-02-24 09:06:19.436411+00:00 [debug] <0.2377.0> Terminating <31028.1894.0> since <31122.2516.0> becomes active rabbit_fifo_dlx_worker
2026-02-24 09:06:19.437003+00:00 [debug] <0.2377.0> Terminating <31122.2516.0> since <0.2600.0> becomes active rabbit_fifo_dlx_worker
2026-02-24 09:06:19.437107+00:00 [warning] <0.2600.0> Failed to process command {dlx,{checkout,<0.2600.0>,2}} on quorum queue leader {'%2F_single_dlx_worker_source',
2026-02-24 09:06:19.437107+00:00 [warning] <0.2600.0>                                                                                 'rmq-ct-cluster_size_3-1-28000@localhost'} because actual leader is {'%2F_single_dlx_worker_source',
2026-02-24 09:06:19.437107+00:00 [warning] <0.2600.0>                                                                                                                                                      'rmq-ct-cluster_size_3-3-28144@localhost'}.

In this case a quorum queue could end up without a dlx worker.

How?

This commit supersedes #15548.

In this commit, we use ra:pipeline_command/4 with a selective receive instead of ra:process_command/3 for the dlx checkout command. This prevents Ra from automatically redirecting the checkout command to a new leader if a failover happens while the command is being processed.

---

This is an automatic backport of pull request #16203 done by [Mergify](https://mergify.com).