libtest: expose --fail-fast as an unstable command-line option

[sourcefrog]

This exposes the fail_fast option added in #105153 on the test harness command line, so that workflows that only want to know if any test fails can find out without waiting for everything to run. For example, cargo-mutants just needs to know if any tests fails. It only works with -Zunstable-options.

(Probably I should make an issue to stabilize it but I just wanted to see if the core change is OK, first.)

[rustbot]

r? @SparrowLii

rustbot has assigned @SparrowLii.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

[sourcefrog]

I can not yet find any in-tree tests for the libtest command line options...

[sourcefrog]

I locally tested this with rustup toolchain link, and the basic function works as intended: after one test fails, no other tests start.

However that does find a problem that doctests don't accept this option.

[jieyouxu]

cc @rust-lang/testing-devex
r? libs-api (libtest cli)

[epage]

Over at rust-lang/testing-devex-team#5, we're building up a list of all of the problems with the current doctest architecture.

I've noted this for the next testing-devex meeting.

[epage]

As this builds on the original implementation and it was recognized as a hack by both the reviewer and the implementer, it would be good to get their feedback. My main concern is I've found PRs a poor place for collecting requirements / design information because they can come and go as different people work on it or even with the same author. Unsure if we should go ahead and do that discussion here or open an issue.

[jieyouxu]

(It's probably best as its own issue, PR discussions are very easy to become lost)