appkit: improve API interaction with InputMethodKit.

[ShikiSuen]

This continues the accidentally-closed #4398 , replacing #4396.

closing #3925 #3814

1. On macOS, `NSTextInputClient::validAttributesForMarkedText` now automatically attaches a value
  called `_rust_winit_\(CARGO_PKG_VERSION)`.

  This allows IME devs to identify whether `winit` is used in the IMKTextInput client.
  If identified, IME devs can introduce compatibility behaviors against such IMKTextInput client.
  InputMethodKit casts `NSTextInputClient` to `IMKTextInput`, but `validAttributesForMarkedText`
  is shared between these two protocols and can all be parsed as `Array<NSString>`.

  `validAttributesForMarkedText` specifically returns an `NSArray<NSAttributedStringKey>` which
  are plain NSString*; Apple treats custom keys as legal, so slipping harmless metadata like
  `_rust_winit_\(CARGO_PKG_VERSION)` in there is the one slot where such data is piggybackable.
2. On macOS, route keyDown through `NSTextInputContext::handleEvent` before raw dispatch,
  suppress stray ASCII from keyUp, and drop stale raw characters when IME commits transformed text
  (e.g. "." -> "。"), aligning character delivery with Cocoa and third‑party IMEs.

• Tested on all platforms changed // cargo +nightly test -p winit-appkit only.
• Added an entry to the changelog module if knowledge of this change could be valuable to users
• Updated documentation to reflect any user-facing changes, including notes of platform-specific behavior
• Created or updated an example program if it would help users understand this functionality // Only tested with cargo +nightly fmt ; cargo +nightly build ; cargo +nightly run -p winit --example ime

The true reason for the 1st modification in this PR (as listed above) is to allow IME devs to automatically enable their customized popup composition buffer window in lieu of preedit (inline composition buffer). The piggybacked metadata contains the version information of the winit package so IME devs can identify buggy winit releases in their IMEs and do their individual accomodations. If this metadata is missing or not detected, then IME can use client bundle identifiers instead (which is comparatively more difficult, still.).

[ShikiSuen]

Unit test logs on my mac:

~/Repos/winit> cargo +nightly test -p winit-appkit
    Finished `test` profile [unoptimized + debuginfo] target(s) in 0.04s
     Running unittests src/lib.rs (target/debug/deps/winit_appkit-787491e69be708a2)

running 6 tests
test app::tests::test_override ... ok
test app::tests::test_custom_class ... ok
test monitor::tests::invalid_monitor_handle ... ok
test monitor::tests::from_invalid_id ... ok
test monitor::tests::monitorhandle_from_zero ... ok
test monitor::tests::uuid_stable ... ok

test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s

   Doc-tests winit_appkit

running 3 tests
test winit-appkit/src/lib.rs - EventLoopBuilderExtMacOS::with_default_menu (line 515) ... ok
test winit-appkit/src/lib.rs - EventLoopBuilderExtMacOS::with_activation_policy (line 493) ... ok
test winit-appkit/src/lib.rs - (line 20) ... ok

test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.04s


running 1 test
test winit-appkit/src/lib.rs - WindowExtMacOS::is_document_edited (line 150) ... ignored

test result: ok. 0 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 0.00s

all doctests ran in 0.65s; merged doctests compilation took 0.20s

[kchibisov]

I can not comment much whether it'll work or not, so some general comment regarding general code comprehension.

Also, have you able to verify that it actually solves the issue and doesn't affect regular Qwerty input (including hotkey handling) and that output is pretty much the same?

[ShikiSuen]

@kchibisov I rebased the commit history of this PR so I can separate the out-of-topic unit test fixes to another PR.

Changes against winit-appkit/src/view.rs now have inline comments. See 16560cf.

[ShikiSuen]

I found a new way to test whether ASCII inputs are handled: cargo +nightly run -p winit --example ime.

Test results showing that some extra fixes are needed. I'll push the fix later.

[ShikiSuen]

This PR is defibrillated now (see e47d639 ). Everything is green. Tests in this video shows that ASCII inputs and the Sogou-style IMEs (Sogou, WeType (WeChat IME), RIME Squirrel) are all passed with no failure of committing the right punctuations / characters.

ScreenRec.2025-11-05.23.44.36.mp4

Note: On macOS we must assume that IME has nothing "disabled". This PR uses Ground state for all input methods, incl. macOS built-in ASCII inputs (ABC or EN-US).

[jrmoulton]

I haven't reviewed the code but I've tested these changes. Regular text input works but while the IME is active modifier keys do not work so as is this is still a breaking change.

(Tested using dev branch of Floem ui library here.)

[ShikiSuen]

@jrmoulton Thanks for your involvement.
Please tell me your modified key combinations. I'll try troubleshooting this on Thursday.
My timezone is GMT+9 and it's my sleeping time now.

[ShikiSuen]

@jrmoulton Please also test 182250d3. I just pushed that.

[jrmoulton]

It is now working perfectly for me

[ShikiSuen]

I hate "merge-commits". I rebased the PR to let them merged before my commits in this PR. Now this PR's HEAD is still the same as what @jrmoulton tested recently.

[jrmoulton]

Can confirm that these changes work for my use cases

[ShikiSuen]

@kchibisov I guess my works are all done here unless you found anything I need to amend.

[kchibisov]

I'd need to test in somehow(don't own macOS) before merging, or @madsmtm can probably take a look.