Deadkeys 2 (#20612)

Re-land of #20515 with less brokenness

In particular it turns out that for control, the .characters() method
returns the control code. This mostly didn't make a difference, except
when the control code matched tab/enter/escape (for
ctrl-y,ctrl-[/ctrl-c) as we interpreted the key incorrectly.

Secondly, we were setting IME key too aggressively. This led to (in vim
mode) cmd-shift-{ being interpreted as [, so vim would wait for a second
[ before letting you change tab.

Release Notes:

- N/A

crates/gpui/src/platform.rs

@@ -688,6 +688,11 @@ impl PlatformInputHandler {
             .flatten()
     }
 
+    #[allow(dead_code)]
+    fn apple_press_and_hold_enabled(&mut self) -> bool {
+        self.handler.apple_press_and_hold_enabled()
+    }
+
     pub(crate) fn dispatch_input(&mut self, input: &str, cx: &mut WindowContext) {
         self.handler.replace_text_in_range(None, input, cx);
     }
@@ -785,6 +790,15 @@ pub trait InputHandler: 'static {
         range_utf16: Range<usize>,
         cx: &mut WindowContext,
     ) -> Option<Bounds<Pixels>>;
+
+    /// Allows a given input context to opt into getting raw key repeats instead of
+    /// sending these to the platform.
+    /// TODO: Ideally we should be able to set ApplePressAndHoldEnabled in NSUserDefaults
+    /// (which is how iTerm does it) but it doesn't seem to work for me.
+    #[allow(dead_code)]
+    fn apple_press_and_hold_enabled(&mut self) -> bool {
+        true
+    }
 }
 
 /// The variables that can be configured when creating a new window