Add cargo_warnings config to control the use of the cargo warning instruction

[scootermon]

Closes: #916

I couldn't come up with a clean abstraction that doesn't needlessly bloat the code so I decided to keep it simple. Because of the lack of abstraction I'm somewhat concerned about maintainability though, as it might become harder to keep track of the warning instructions.

I also found some instances of println!s, which looked to me like they're supposed to be using the cargo warning instruction. I adjusted those, but let me know if that's incorrect.

[scootermon]

This is only here because the new nightly versions correctly detect that this field is never used, which causes the CI to fail.

[NobodyXu]

Thank you for the PR!

Can you please add a test for which warnings are disabled?

[NobodyXu]

@scootermon if it's too hard to add testing, I'm ok with only test them on unix (since tests failed on windows)

[scootermon]

@scootermon if it's too hard to add testing, I'm ok with only test them on unix (since tests failed on windows)

I added testing in the latest commits and there shouldn't be any problems on Windows because I'm not actually forking the process. The CI is failing because the tests are working :) We're still writing the warning "Compiler version doesn't include clang or GCC" to the console (because that's handled in Tool not in Build).
I took some time off to think about how to deal with it but I would like to hear your input on it:

1. allow specific warnings to be written in the unit test
2. pass down the cargo_warnings flag into Tool to suppress it. Not as clean because we can't use the Build::print_warning method.

[NobodyXu]

I suppose we should pass down the cargo_warnings flag to `Tool.

I suppose we can add a new type:

struct CargoWarning(bool);

impl CargoWarning {
    fn print_warning(&self, msg: &dyn Display) { .. }
}

So that the print_warning can be shared between Build and Tool.

[scootermon]

@NobodyXu, apologies for the delay, got caught up with other stuff.

I did end up having to skip the warnings_on test case on MSVC though because cl.exe apparently doesn't write compilation warnings to stderr...

[NobodyXu]

Thanks, LGTM

Can you also resolve the conflict please?

[NobodyXu]

Thank you!