Window and webview commands are all async functions

[Legend-Master]

Most of if not all of window and webview commands are sending the command to main thread and handled there, adding an async layer on top of it feels wrong to me, also this increases the binary size by 300-400 KB (tested with release build of helloworld example on Windows 3.60MB -> 3.96MB)

tauri/crates/tauri/src/window/plugin.rs

Lines 49 to 79 in fdaf48f

	macro_rules! getter {
	($cmd: ident, $ret: ty) => {
	#[command(root = "crate")]
	pub async fn $cmd<R: Runtime>(
	window: Window<R>,
	label: Option<String>,
	) -> crate::Result<$ret> {
	get_window(window, label)?.$cmd().map_err(Into::into)
	}
	};
	}
	macro_rules! setter {
	($cmd: ident) => {
	#[command(root = "crate")]
	pub async fn $cmd<R: Runtime>(window: Window<R>, label: Option<String>) -> crate::Result<()> {
	get_window(window, label)?.$cmd().map_err(Into::into)
	}
	};
	($cmd: ident, $input: ty) => {
	#[command(root = "crate")]
	pub async fn $cmd<R: Runtime>(
	window: Window<R>,
	label: Option<String>,
	value: $input,
	) -> crate::Result<()> {
	get_window(window, label)?.$cmd(value).map_err(Into::into)
	}
	};
	}

I wonder why we did this, since pretty much everything else uses normal functions

I'm aware of the dead lock I mentioned in #10210 (comment), if that's the reason we did this, I feel like probably should fix that instead

Edit: did a bit more digging, seems like the dead lock is caused by us locking the plugins for invoking commands, and during the invoke, the command sends an event to the event loop synchronously and the event handler also needs to lock plugins for the event listeners, not sure why on_event and extend_api take mut self though