Add a second worker that runs all GoodJob queues. The "base" worker will
run all queues except the :auxiliary queue.

This, combined with the high priority assigned to all jobs in the
:auxiliary queue, effectively builds a fast lane; all jobs run in the
:auxiliary queue will be executed before all other jobs by the auxiliary
worker.

The fact that the base worker does not run any jobs in the :auxiliary
queue means we can have this fast lane without putting every other job
on hold for, potentially, long periods of time.

P.S. The reason we use `before_enqueue` in ApplicationJob is because
GoodJob does not support setting a default priority per queue.

This resolves the "duplicate name" warning when deploying. However, it
may remove all other ENV vars from the base worker, so this change may
not stick around

This is an attempt to resolve this helm upgrade error:
> The order in patch list: ... doesn't match $setElementOrder list: ...

Attempting to merge/override the same ENV variable in different
deployments ended up being very difficult. This is perhaps less elegant,
but simpler

Attempting to use the --queues flag didn't work, but using the ENV
variable does