Multiprocess backend

I got curious and started off a impl that could separate the backend from the main process if we wanted to investigate a multi-process design in the future. The idea being that if we hit a situation like the IRC bridge where one process can actually get quite busy, then it would make sense to be able to split the load.

This turned out to be more work than I could comfortably commit to, but we should definitely try to support this in the future for large deployments.

The branch is https://github.com/matrix-org/matrix-bifrost/tree/hs/multiprocess

Start by reviewing the hs/multiprocess branch linked in the issue and comparing it with the current matrix-bifrost codebase. Determine the intended process boundary and deployment behavior before attempting work. Done would mean a supported multiprocess backend design suitable for large deployments, with the scope and validation criteria made explicit.

Written by the indexing model from the issue text.