Should Datagen Options be Copy?

[sffc]

We have a policy of strongly preferring Options structs to be Copy. But the new Datagen Options contain some HashSets. Should they instead be borrowing those fields?

[sffc]

https://unicode-org.github.io/icu4x/docs/icu_datagen/options/struct.Options.html

• @Manishearth - Datagen is its own thing. When we decided the policy of Copy I was mostly thinking about components. That's not to say that we shouldn't make this Copy, but I feel it's out of scope of the general policy.
• @robertbastian - Datagen is special and not panic-free.
• @sffc - Datagen has good reasons for all the places it differs from the components style guide. It's not clear why this case rises to that level.
• @sffc - Borrowing allows the reuse of options, which seems useful
• @sffc default-spread for non-exhaustive will probably require Copy
• @robertbastian - borrowed ergonomics are worse
• @Manishearth - The component styles were designed for reasons specifically about components. But we can look at these two reasons. For re-use of the options, it's not clear if that's compelling for Datagen. For default-spread, that seems valid. It's not super conventional for options structs to get lifetimes. I'm not opposed to making it Copy, but I'm not convinced we must.
• @robertbastian - This might not need to be an options bag. Where does the export method live? It could become a free function. It could also be a method on the options bag.
• @sffc - I do think we should move in the direction of a DataExporter that owns these options and then plumbs a DataProvider to a DataConsumer. And that decision changes whether or not we borrow the options.

Decision: Table this discussion until #3564 is resolved.