Performance runs can hang for 4 hours blocking machines in the helix queue #38136
Description
Currently we have long timeouts in the perf workitems 4 hours.
While I understand this for the dotnet/performance repo since it does a bunch of iterations to get the perf numbers, in PR builds only 1 iteration is made. So we should be able to override the WorkItems timeout to 15 mins when we're on a PR or CI build outside of the performance repo.
Also, we should use the path filtering mechanism we have for the runtime pipeline to run coreclr only when coreclr or libraries is changed and mono only when mono or libraries is changed.
This is not just affecting dotnet/runtime but also affecting other repos which rely on the same helix queue, which we're now blocking.
cc: @dotnet/runtime-infrastructure @DrewScoggins @MattGal @billwert @adamsitnik