test: fix flaky child-process-fork-regr-gh-2847
Windows would die with ECONNRESET most times when running this particular test. This commit makes handling these errors more tolerable.
I'm not sure if the error handling logic is 100% correct here. Is silencing ECONNRESET errors on server-side sockets ever acceptable, or are those particular errors relevant to this test?
/cc @mhdawson @jasnell @indutny @trevnorris