feat: initialize package

[ceholden]

Description

This PR adds an updated "ECS protection manager" that I originally wrote to help prevent scale-in on long running ECS tasks that are auto-scaled based on queue depth. Once the worker count scales up and they start consuming tasks, the message depth will go down and trigger scale-in. If the tasks are doing work we want to prevent this from happening!

I added some features to mature this helper,

1. Debounce API calls to avoid rate limits
2. Use a file lock to prevent parallel workers from releasing task protection (e.g., Celery running with --concurrency=4 workers won't override the task protection from other workers)
3. Support injecting the protection manager context to give the tasks control over the task protection. If the task is running long then it can extend the task protection (e.g., if the task is somewhat unbound in runtime)