Support max_concurrency for async APIs

[deliahu]

Description

Currently, async APIs process one request at a time (i.e. max_concurrency is 1). Support the user specifying max_concurrency > 1, which will result in concurrent requests into the container.

Notes

One approach is to have multiple goroutines, all pulling from the SQS queue. Another option is to have a single goroutine pulling more than one message at a time from the queue, and then distributing the messages across goroutines (care will have to be taken to avoid idle goroutines).

Misc

• Update async/autoscaling.md and async/configuration.md accordingly (add new sections, update default values for target_in_flight and max_concurrency, etc)

• Here is the commit which removed max_concurrency from the async docs: 7508c76