Reduce the number of text-wrap features to just 3

[foolip]

Only text-wrap: balance and text-wrap: pretty are significant new
features, let everything elese be covered by the generic text-wrap
feature.

This changes the Baseline date of text-wrap a little, but it's still
in 2024 and so it doesn't seem worth trying to pin it to the earlier
date by trimming the compute_from list in a way that wouldn't match
the feature decription.

Fixes #2543.

[ddbeck]

I'm happy for this to happen at some point, but wiping out features is a semver major event. I don't really want to do that as a one-off for this. So either this is blocked pending a resolution for at least #91 and a resolution on as-yet un-described "inferior feature" capability (i.e., the structure a merged feature becomes).

[foolip]

@ddbeck what's the "inferior feature" capability you have in mind? Here, I'm imagining that all removed features are simply redirected to the merged text-wrap feature.

[ddbeck]

Philip and I chatted about this today and he made an OK argument for the idea that these aren't real merges, but actually a kind of error correction (i.e., we created features that didn't/shouldn't really exist and can use redirects as proposed in #91 instead of something richer/more complex, at least initially).

Given that context, I said I'd look at this again. I had a hard time following all the moving parts here, so I drew a diagram. I think this shows how things are moving around here:

flowchart LR
    text-wrap
    text-wrap-balance
    text-wrap-mode
    text-wrap-nowrap
    text-wrap-pretty
    text-wrap-stable
    text-wrap-style

    text-wrap-stable --> text-wrap

    text-wrap-mode --> text-wrap

    text-wrap-nowrap --> text-wrap

    text-wrap-style -- text-wrap-style: auto
    text-wrap-style: nowrap --> text-wrap
    text-wrap-style -- text-wrap-style: balance --> text-wrap-balance
    text-wrap-style -- text-wrap-style: pretty --> text-wrap-pretty

Loading

I can see Philip's point here. I'd prefer the purist (and purest) version, but simple redirects could work here. I suggest we come back to this as soon as #91 is resolved, which will hopefully be soon.