The catchall kubernetes shard should listen to only the default and mailers queues in production

Having made application changes in #987 (closed), we can make the production config changes. We are focused here on kubernetes catchall, as that shard handles more workers, and also handles new workers that are rolled out.

Setup

We start with a couple of do-nothing items:

1. Use the exclude_from_gitlab_com tag as a way of testing routing rules in staging and production, with zero production impact: #1072 (closed)
2. Add the routing rules for all the other shards, but with null as the target (so they do nothing). This makes the following steps simpler: #1073 (closed)

Migrate

1. Pick some initial batches of queues by feature category and migrate those: #1074 (closed)
2. Migrate the rest and stop listening to all the other catchall-on-k8s queues, hopefully saving us some CPU! #1075 (closed)