Kill IMAP processing chain

Today we create a processor chain to manage IMAP request: each processor passes the request to the next until the request is managed.

This leads to absurdly, needlessly long call stacks as we easily have 40+ processors.

• This have a performance impact as feeling thread stacks do not come for free
• Exceptions are hard to read (longer stacktraces)
• Reading flame graphs is also harder!

We should in my opinion just use a map to directly get the processor for a given request, simply using the request class as a key.