test: fix flaky test-cluster-shared-leak

[claudiorodriguez]

Pull Request check-list

Please make sure to review and check all of these items:

• Does make -j8 test (UNIX) or vcbuild test nosign (Windows) pass with
  this change (including linting)?
• Is the commit message formatted according to CONTRIBUTING.md?

NOTE: these things are not required to open a PR and can be done
afterwards / while the PR is open.

Affected core subsystem(s)

test

Description of change

Please provide a description of the change here.

Test was flaky on centos7-64 due to an uncaught ECONNRESET on the worker code. This catches the error so the process will exit with code 0.

Fixes: #5604

PS: I did consider retrying but ECONNRESET was already being disregarded on the cluster-master side of things, and the test itself is concerned with a particular assertion in cluster, but I'd still like some feedback on this.
You can reproduce the error by adding a timeout like this:

//...
conn.on('error', function(e) {
  // ECONNRESET is OK
  if (e.code !== 'ECONNRESET')
    throw e;
});
// add this and run the test a couple times:
setTimeout(function(){
  conn.destroy();
}, 5);
//...

[claudiorodriguez]

P.P.S: not the first time centos has this kind of behaviour on the CI, there's several tests that had "flaky" behaviour because of prematurely closed connections in that platform. That probably warrants further investigation.

[claudiorodriguez]

CI: https://ci.nodejs.org/job/node-test-pull-request/1970/

[Trott]

LGTM

[Trott]

@nodejs/testing

[Trott]

Confirmed that this still catches the error it's supposed to catch in Node v4.2.1.

[santigimeno]

LGTM.
I've been able to reproduce the failure in OS X by stress testing it with nodejs/testing#3 (comment). The proposed patch fixed the error.

[jasnell]

LGTM

[jasnell]

I assume this is relevant for v4 also, please remove the lts-watch label if not! thanks!

[jasnell]

Landed in 74a703d