Add logic to Duration::total to handle loss of precision

[nekevss]

This adds the general logic from FractionToDouble to handle edge case Duration::total operations.

We still need to figure out a good way to test this in Rust, but all the Duration suite tests pass when ran locally in Boa.

[Manishearth]

Shouldn't be hard to just write the existing tests in Rust.

[nekevss]

I have one already from another branch, but it has a bug and is creating the wrong expected value. So I need to clean it up a bit before adding it.

[Manishearth]

I pushed a test that seems to work. I used the expected values from the JS tests that were failing.

[nekevss]

Ah, that will work. I was hoping to have a full version of the test262 test. But I'm not sure it's going to be possible without bringing in boa_string as a dev-dependency for access to a JsString type.

[Manishearth]

Ah, that will work. I was hoping to have a full version of the test262 test. But I'm not sure it's going to be possible without bringing in boa_string as a dev-dependency for access to a JsString type.

Wouldn't be a good test then IMO since that behavior itself could be buggy.

[nekevss]

Wouldn't be a good test then IMO since that behavior itself could be buggy.

It would be the best representation of the test262 test in Rust. But I don't necessarily disagree.

In any case, we have at least a few engines running the test262 suite to help flag potential regressions. It would be nice to catch it in our tests.