Add priority queue

[donaldgray]

Adds a new endpoint: POST /customers/{customer}/queue/priority (mimics POST /customers/{customer}/queue but skips queue and ends up in priority DLCS queue).

A priority submission is expedited both in terms of processing and ingest:

• Process: tasks are enqueued to a dedicated queue/cluster (named by PRIORITY_QUEUE_NAME) with its own small worker pool, so they are never blocked by a standard-queue backlog. Each engine container runs a second qcluster process for it (entrypoint-worker.sh);. This means no new instances or services running as I expect use to be rare and want to avoid infra overhead.
• Ingest: batches for priority collections are POSTed to the DLCS /customers/{customer}/queue/priority endpoint, so they also skip any DLCS-side backlog.

If PRIORITY_QUEUE_NAME is unset the behaviour is as before

Commits

1. django-q==1.3.9 → django-q2==1.10.0, the maintained drop-in fork. Prerequisite for multi-queue routing (ALT_CLUSTERS), also seems like sensible update regardless.
2. Add priority queue for expedited submissions - the feature itself: new route, Collection.priority field + migration, cluster routing in the view, DLCS priority ingest, worker entrypoint. Also repaired the integration test stack, which referenced an old Dockerfile.CompositeHandler left over from when this was part of the protagonist.

Configuration

New configuration:

Variable	Default	Notes
PRIORITY_QUEUE_NAME	unset (disabled)	With the SQS broker, must equal the name of a real SQS queue
PRIORITY_WORKER_COUNT	1	Priority workers per engine instance

Infrastructure

New priority SQS queue (same settings as the standard one) and set the above env vars on the API and engine services.

Migrations

New migration to add priority field.