This fixes a flake on linux CI which started recently where the "write"
promise is being rejected after the request has been aborted /
cancelled. In this case we should drop the error to the floor but
instead we pass it down the stack where it eventually emits a now
unhandled error event.
Example failure: https://app.circleci.com/pipelines/github/electron/electron/38072/workflows/c1faf19b-aa41-4f99-a564-165729222859/jobs/838813
Verified fix by running the test that caused it 10000 times before fix
and 10000 times after. ~50 failures before, 0 after.
* fix: allow net streams to be throttled [WIP]
* fix: handle resume throttling within IncomingMessage [WIP]
* fix: fix urlLoader typing, add throttle test
* fix: fix lint and increase test timeout for Linux
* fix: increase test chunk limit to 20 and timeout to 2000
Co-authored-by: Jeremy Rose <nornagon@nornagon.net>
