Clearing the checkbox makes the processor thread delete the directory before running, which is just really fuckin' spooky if you don't know that and accidentally set the string to something bad like the current dir! Ideally we should have a proper button which just enqueues a command for minidump-processor.

This requires a change to minidump-processor to give it a proper "queue" of commands because it will be the first command that is "TCP-like" and not "UDP-like". That is, currently there is only one buffered command because all the current commands basically obsolete whatever the processor is currently doing (hence why cancellation of the stackwalk is done by every command).

I think we do want some amount of cancellation though, so maybe there should be a queue of TCP-like commands and a single UDP-like command. Alternatively it might be desirable to also queue up the UDP-like ones too so that the processor is forced to ack-back even if it "notices" that the commands are obsoleted.

There is currently some mild racey-ness in the code where I believe in principle you could get the UI confused about which partial-processed report it's doing incremental updates too, although I don't think this is a very serious concern in practice, as it would require sending a lot of different commands in incredibly fast succession, and having the processor grab an update in the middle of that.