CI: don't block further CI results on tidy/fmt

[CAD97]

Some irlo context.

Currently, if a PR has even trivial tidy/fmt errors, the tidy/fmt error early-exits the PR smoke test CI and reports just the tidy/fmt error.

This can be problematic because of the long test cycle; it would reduce the number of required pushes/CI runs if CI could report both tidy/fmt nits alongside the test harness state.

[Mark-Simulacrum]

We can likely move tidy to the mingw-check builder and skip it on the longer builder.