Special case testdriver.js when detecting affected tests

[Hexcles]

IIRC, testharness.js is treated specially in the heuristics for detecting affected tests (whenever it's modified almost all tests are affected). I think we need to do the same for testdriver.js as its use increases.

[foolip]

@LukeZielinski can you take a look at this? Does touching testdriver.js currently cause stability checks to time out, or how much margin do we have?

A similar case is idlharness.js where @lukebjerring added code to run more affected tests, but in that case we don't expect the number of tests to ever go much beyond the total number of specs/directories.

[lukebjerring]

#13392 is the change mentioned

[LukeZielinski]

So empirically (#16852) the stability checks were not run for a no-op change to testdriver.js. Although it's kind of surprising, since (as you said) only resources/testharness* is excluded in testfiles.py.

Is there some other configuration at play here?

[foolip]

@jgraham do you know why touching testdriver.js won't end up running any tests? I'm pretty sure it once did, and I haven't changed it or been aware of a change.

[foolip]

Ping @jgraham on #15485 (comment).

[gsnedders]

@foolip they were in #20674; no idea why they weren't in #16852.

[stephenmcgruer]

Marking roadmap due to inactivity (and lack of concern about inactivity)