test: fix flaky test-inspector-port-zero-cluster

[Trott]

With a properly functioning test, it is possible for a cluster worker to
fail to launch due to a port collision. For better or for worse, this is
working as expected and so the test now accommodates that reality.

Fixes: #13343

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

Affected core subsystem(s)

test inspector cluster

[Trott]

stress test against master (should show some failures): https://ci.nodejs.org/job/node-stress-single-test/1304/

[Trott]

stress test against this PR (should show no relevant failures): https://ci.nodejs.org/job/node-stress-single-test/1305/

[Trott]

CI: https://ci.nodejs.org/job/node-test-pull-request/8683/

[Trott]

ping @bnoordhuis

[refack]

Hey! That's my trick #13373...
(anyway IMHO test should be simplified to just parse argv anyway, as that's actually the unit of code it's validating)

[Trott]

Stress test shows expected failure on master at https://ci.nodejs.org/job/node-stress-single-test/1304/nodes=debian8-64/console. No failures on the stress test CI for this PR.

[Trott]

Hey! That's my trick #13373...

@refack Ah, yes, similar for sure. Had I realized they were that similar, I would have suggested that the fix for the flaky test be split out into its own PR so that it doesn't get delayed by reviews for the new tests and any other refactoring that might be going on in that PR. In fact, not too late to do that if you'd prefer.

[Trott]

(Although I now see that I said I was +1 on creating a known issues test at the same time, so clearly I can't be trusted.)

[Trott]

CI again since I landed a couple of minor to-be-squashed commits since the last CI: https://ci.nodejs.org/job/node-test-pull-request/8686/

[refack]

Hey! That's my trick #13373...

@refack Ah, yes, similar for sure. Had I realized they were that similar, I would have suggested that the fix for the flaky test be split out into its own PR so that it doesn't get delayed by reviews for the new tests and any other refactoring that might be going on in that PR. In fact, not too late to do that if you'd prefer.

I just saw that the blocking review has been dismissed, so I'm going to land #13373 ...

[refack]

Killed https://ci.nodejs.org/job/node-stress-single-test/1304/nodes=pi1-raspbian-wheezy/
and https://ci.nodejs.org/job/node-stress-single-test/1305/nodes=pi1-raspbian-wheezy/
That were running for 10 hours

[Trott]

#13373 landed so this is unnecessary.