Triaging a reftest from the test page links the failure to a "Harness status" subtest

[foolip]

Steps to reproduce:

1. Start at https://wpt.fyi/results/css/css-contain/contain-size-monolithic-002.html?run_id=5750172413853696&run_id=5751886844002304&run_id=5730426872135680
2. Enable triage mode
3. Select the red cell for the Chrome failure
4. Click "triage"
5. Link https://bugs.chromium.org/p/chromium/issues/detail?id=1154574 and submit

The result is https://github.com/web-platform-tests/wpt-metadata/pull/670, and this bit:

https://github.com/web-platform-tests/wpt-metadata/blob/abcde9ed320b7428bf728683c214304666706be0/css/css-contain/META.yml#L45-L49

This has happened once before and I noticed + clean it up in https://github.com/web-platform-tests/wpt.fyi/issues/1958#issuecomment-715960474, but didn't file a new issue then.

("Harness status" is not the right terminology here, but that's tracked in https://github.com/web-platform-tests/wpt.fyi/issues/1958. If fixed, presumably we'd instead get bogus subtests called "Test status" in the triage metadata.)