Flexbox: don't wrap auto-height column containers against ancestor-derived available space

[nicoburns]

Objective

Fixes #1175.

An auto-height flex-wrap: wrap column container could wrap its lines against AvailableSpace::Definite main-axis space handed down by an ancestor (typically during a ComputeSize/ContentSize measurement round), even though its own main size is content-based. This showed up as one-frame layout flicker while animating a child's height: for the frames where the content total exceeded the ancestor-derived available height by a couple of ulps, the container split into two columns. Chrome keeps a single line for the equivalent DOM.

The wrap constraint in collect_flex_lines (and collect_balanced_flex_lines) previously fell back to available_space.main(dir) whenever the container had no max main size:

let main_axis_available_space = match constants.max_size.main(dir) {
    Some(max_size) => AvailableSpace::Definite(/* min(available, max) when size indefinite */),
    None => available_space.main(dir),   // <-- ancestor-derived Definite leaked in here
};

Now, for a column container with no definite main size and no max main size, definite available space is replaced with MaxContent (no wrapping), since an automatic block-axis size never resolves against available space:

    None if !dir.is_row() && !has_definite_main_size && available.is_definite() => AvailableSpace::MaxContent,
    None => available_space.main(dir),

Rows are unchanged: an automatic inline-axis size does resolve against available space (fill-available), so row containers still wrap against definite available width. Min/max-content constraints and containers with a definite main size or a max main size are also unchanged.

Context

• Issue Auto-sized wrapped flex container wraps lines against ancestor-derived definite available space during content-size measurement #1175's repro (persistent tree, per-frame set_style + compute_layout_with_measure) produced 2 anomalous frames on main; 0 after this change.
• PRs Flexbox: use hypothetical main sizes for column intrinsic contributions #1165 and Flexbox: match Chrome's row intrinsic sizing (clamp order, multi-line min-content, line re-collection) #1166 do not fix this issue (verified by running the repro against both branches) — they change intrinsic contribution sizing, while this bug is in line collection. This fix also cherry-picks cleanly onto Flexbox: match Chrome's row intrinsic sizing (clamp order, multi-line min-content, line re-collection) #1166 and fixes the repro there.
• New gentest fixture wrap_column_auto_height_does_not_wrap_against_available_space: a non-stretched auto-height column-wrap container (2×100px items) inside a 150px-tall row parent. Chrome-generated expectation is a single 60×200 column; on main all 4 generated variants fail (taffy wrapped into two columns), with this change all 6069 generated tests pass (default and --all-features).
• CHANGELOG updated.

Feedback wanted

The fix keys on !dir.is_row() as a proxy for "main axis is the block axis" (taffy has no writing modes). If you'd prefer the same treatment for rows under some condition (e.g. only during ContentSize sizing rounds), happy to adjust — but as-is, rows keep the fill-available wrapping behavior that matches browsers for block-level row containers.

Link to Devin session: https://dioxus.staging.devinenterprise.com/sessions/b637316906704d10ba77a4da9f248d4e
Open in Devin Desktop: https://dioxus.staging.devinenterprise.com/desktop/session/b637316906704d10ba77a4da9f248d4e?variant=devin-insiders
Requested by: @nicoburns

[staging-devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR that start with 'DevinAI' or '@devin'.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring