Provide a getter for the resolved options for the collator

[hsivonen]

Fixes #3880

(Before merging, we should decide if we want the return type to be CollatorOptions of if we want a ResolvedCollatorOptions that doesn't wrap each field in Option.)

[hsivonen]

It looks like CI wants the Diplomat wrapper to go into the same PR.

I'm converting this to draft until it's been decided whether this should return CollatorOptions or a new ResolvedCollatorOptions that doesn't wrap each field in Option.

[sffc]

Discuss with:

• @zbraniecki
• @sffc
• @younies

[zbraniecki]

In my prototype of icu_preferences I moved in the direction of a macro generating resolved options without Option. I think this is the direction we should take here.

@hsivonen are you opinionated in the other direction?

[hsivonen]

In my prototype of icu_preferences I moved in the direction of a macro generating resolved options without Option. I think this is the direction we should take here.

@hsivonen are you opinionated in the other direction?

I am not. That is, without Option works for me. Thanks.

[sffc]

I'll weigh in here. I think the input and output options should be separate structs.

In ECMA-402, the input and output structs are sometimes not the same. For example, Intl.NumberFormat takes localeMatcher as an input option but locale as a resolved option (though we have other mechanisms in icu4x for dealing with locale matching and supported locales).

But more importantly, the semantics of resolved options are different. The required versus optional fields are not the same. For example, in Intl.NumberFormat, currency is optional in both input and output, but style is always present in output. (This raises an interesting question about whether we should express the dependence between options in the type system so that currency is required when style = "currency"; I think not, just return an error according to the ECMAScript specification.)

This would imply that the options should also be different structs across FFI.

Ideally these FFI structs should also include optional fields, but that is blocked on rust-diplomat/diplomat#132.

[hsivonen]

This would imply that the options should also be different structs across FFI.

I have now made the FFI types distinct from each other even though the types are, for now, identical except for the name.

Ideally these FFI structs should also include optional fields, but that is blocked on rust-diplomat/diplomat#132.

I have not duplicated the enums that go in the FFI option bags, so on the resolved side Auto is guaranteed-unused, but that's not communicated via the type system.

[sffc]

As discussed on 2023-11-09, we're happy with the direction of this PR.

[hsivonen]

Blocked on #4326. It appears that dart provided by the flutter snap doesn't satisfy the formatting expectations of CI.

[hsivonen]

Looks like merge conflicts show up faster than my cycle to check back here for the CI results. I'll merge main again.

[hsivonen]

If/when this get re-approved, please push the landing button, too, so that this doesn't get out of sync with main before I have an opportunity to push the landing button.