Move icu::calendar::week to experimental

[sffc]

Given https://unicode-org.atlassian.net/browse/CLDR-18275, it is unclear that this code in its current state is what we will need. I think we should mark it as experimental.

[robertbastian]

This code is also used to determine the first day of the week, and the weekend. Only the min_week_days field, and the week_of_month and week_of_year methods fall under CLDR-18275.

[sffc]

Yeah. So maybe only move the "bad parts" to experimental.

[robertbastian]

They can't be moved, they're all on the same type. Unless you want to remove the min_week_days field from the data struct, and make a new data struct in experimental which just contains a single u8, and requires the existing data struct to implement week_of_month and week_of_year.

We can just make those methods private if you don't want them to be stable in 2.0

[sffc]

min_week_days is such a small amount of data that I'm not inclined to spend a lot of time on removing it right now.

But I think we should hide the methods behind an experimental feature. Or private, but probably experimental.

[robertbastian]

I don't think this justifies adding an experimental feature. They are not experimental, they work, and they have been stable for multiple releases. If we say they are footguns, we should remove them until a client asks for them.

[sffc]

I don't think this justifies adding an experimental feature.

I think having an experimental feature should be seen as having a very low cost. It isn't something that should factor very highly into our calculus

They are not experimental, they work, and they have been stable for multiple releases.

And CLDR plans to deprecate this functionality in favor of a different algorithm

If we say they are footguns

They aren't footguns; they are soon but not yet deprecated

we should remove them until a client asks for them.

I would approve a PR removing them. But then we should probably also remove the field from the data struct.

[robertbastian]

I think having an experimental feature should be seen as having a very low cost.

Ideally we would have experimental code in icu_experimental, not in mature crates behind experimental features.

They aren't footguns; they are soon but not yet deprecated

CLDR is talking about deprecation because they are footguns:

Research, posted in CLDR-17095, suggests that users are surprised when changing their locale changes their week numbering.