Enable reftypes tests on aarch64.

[cfallin]

With #1852, we now have support for reference types in the new backend
framework generally, and on aarch64 in particular. This PR removes the
test-ignore directive for reftypes tests on this platform.

This actually enables reftypes support for all platforms, rather than explicitly
only on x86_64, but going forward all new backends should include reftypes
support (since the new-backend framework does, and it just takes a few
machine-specific opcodes beyond that to do so), so I think it's best just to
set this baseline now.

Fixes #1886

[alexcrichton]

There's an instance of #1886 in tests/all/main.rs, mind removing those #[cfg]? I think this can close #1886 as well afterwards?

In any case, very nice!

[fitzgen]

There's a few of them now, I think all of these cfgs can be removed:

tests/all/func.rs:126:#[cfg(target_arch = "x86_64")]
tests/all/func.rs:449:#[cfg(target_arch = "x86_64")]
tests/all/func.rs:484:#[cfg(target_arch = "x86_64")]
tests/all/main.rs:23:#[cfg(target_arch = "x86_64")]
tests/all/main.rs:25:#[cfg(target_arch = "x86_64")]
tests/all/main.rs:29:#[cfg(target_arch = "x86_64")]

[fitzgen]

Also, thanks! 🎉 🎉 🎉

[cfallin]

Argh -- false positive: removing the directives in tests/all/main.rs, I see failures of gc::smoke_test_gc and gc::gc_during_gc_from_many_table_gets. I wonder if the tests enabled in build.rs are only testing opcodes and the gc tests exercise the stackwalking more?

I'll push the commit in any case so the failures are visible here on the CI run.

[fitzgen]

It is possible that we don't dynamically trigger any GC during the whole reference types proposal spec tests, and so they would only be testing the code generation.

[cfallin]

Closing this now, as #2410 enables GC/reftypes on all backends!