Currently, lldb-vscode is reading and processing requests one at a time, which
doesn't match exactly the interactivity of the UI. What happens when LLDB is
working hard parsing symbols, and at the same time the user is hovering around,
pressing resume multiple times, expanding variables in the variables view, etc?
In this case, LLDB will eventually process all of these actions one by one
even though the user has already pressed resume, which means that most of the
responses will be discarded by the UI. This imposes a big burden on lldb.
A way to fix that is to parallelize the reading and processing of requests,
so that while a request is being processed, the reader can filter out pending
This diff only contains the concurrent queue used for this. If this code makes
sense, I'll proceed to add some filtering logic for pending requests.