Support separated worker daemons

Requirements...

The 'test publisher' (i.e. parallel_cucumber) makes an ?http service available, through which workers can retrieve the application and test files - probably as zip archives. It publishes its address into the database's 'awaiting workers' table, receives results (again, probably as zips), and unpublishes itself once complete.

Worker-owners monitor the 'awaiting workers' table and somehow coordinate to share resources between 'awaiting workers' tasks, so that if there is only one task, all available suitable workers run it, but resources are evenly and dynamically allocated if additional tasks dis/appear - down to a minimum service level - e.g. min=4 phones. Unclear what happens if several tasks are already at minimum service, and a phone disappears: a task becomes underpowered or is suspended pending workers? Where is the minimum service level defined? What prevents or permits hogs demanding lots more?

Implies worker-owners also register with the database.

'Somehow' suggests a task-resource allocator.

By default, parallel_cucumber will launch its own worker-owner, task-resource allocator, assuming solitary operation and test-execution facilities.

No files, tests, or entry points are identified. Start by mapping parallel_cucumber's test publisher, worker-owner, and database "awaiting workers" table, then clarify resource-allocation rules and minimum service-level behavior before implementation.

Written by the indexing model from the issue text.