Add a container cleanup job to the linux builders

[finestructure]

We've had multiple occasions during the Wasm and Android processing where "stalled" containers blocked processing:

 docker-user@bastion  lb1  ~  docker ps                                                                                                                                                                                      Sat May 31 09:46:36 2025
CONTAINER ID   IMAGE                                                             COMMAND                  CREATED             STATUS             PORTS     NAMES
5252dab38279   registry.gitlab.com/finestructure/spi-images:basic-6.1-latest     "swift build --tripl…"   About an hour ago   Up About an hour             interesting_hamilton
991a521491bf   registry.gitlab.com/finestructure/spi-images:android-6.1-latest   "swift build --swift…"   2 hours ago         Up 2 hours                   musing_hypatia
3e1558adbc95   registry.gitlab.com/finestructure/spi-images:android-6.1-latest   "swift build --swift…"   2 hours ago         Up 2 hours                   recursing_ishizaka
a8a0b51382c2   registry.gitlab.com/finestructure/spi-images:wasm-6.1-latest      "swift build --swift…"   2 hours ago         Up 2 hours                   angry_wilbur

We should have a job running (perhaps as part of the builder itself) that force removes containers older than 15 mins. The only thing we need to be careful about is not to remove deployment containers, as these have a longer timeout. The IMAGE is probably something we'll want to look at for that purpose.

[finestructure]

Adding this to the builder is tempting but potentially problematic: If the linux builder has as many stalled builders as it is wide, there won't be another builder spawned to clean them up.

This will probably need to be a separate process running on each builder.