Cancel writing events after the Kill button is pressed

[sorawee]

When we press the kill button, make it so that queued up "do-insertion"s are not processed.

This code is originally authored by @rfindler

Fixes racket/racket#1677 and racket/drracket#444

[sorawee]

@rfindler this is the only place that writes to output-ports-clear-count, so provided that there is only one thread that calls (clear-output-ports), I think there won't be a race condition?

[capfredf]

I've fixed the CI issue in #282

[rfindler]

There's a race because the set! happens in one thread and the lookup ((define current-clear-count output-ports-clear-count)) happens in a different thread. I think that because racket guarantees sequential consistency for thread (it doesn't for futures or places) that this race condition is benign and probably this is okay, overall.

That said, I feel like it would be better if there were some way to use the existing syncronization that's happening in this code to be able to cancel the pending work.

[sorawee]

Closing in favor of b51090c