Consider adding buffered async output, maybe DETS + gen_stage

Right now the async casting logic _heavily_ assumes short WP.com API outages. Historically, that's fair, but I think we can do better.

It'd also be nice to be more robust against proc brutal kills or unexpected host shutdowns.

Start by locating the async casting logic in the Elixir API client and read how it handles WP.com API outages. The issue suggests DETS and gen_stage, plus resilience to process kills and host shutdowns; done would require an agreed design and implementation covering those failure cases.

Written by the indexing model from the issue text.