Add cosmic_atspi_v1 protocol

[ids1024]

This works with the receive example of reis using LIBEI_SOCKET=/tmp/atspi-ei-kb.socket. I've also worked on changes to at-spi2-core, so I'll now be able to try testing it...

If a version of this is ultimately used, it will need a secure way to pass the socket to accessibility tools. Putting it in /tmp is a placeholder.

[ids1024]

It's just a hacky prototype at the moment, but this now partly works with https://github.com/ids1024/at-spi2-core/tree/atspi-ei.

Orca keybindings now work, but the keys are also sent to applications. We need something like X11's XIGrabKeycode to register a key combination. But all we need to do is have the compositor track those grabs and not send them to applications. We don't need to feed all input synchronously through the screen reader or anything.

I think that's pretty much the only other thing we need for parity with AtspiDeviceX11?

We'll also need a secure mechanism to start the screen reader, and only give it access to this.

[ids1024]

I've fixed an unwrap panic I noticed sometimes when killing orca, and added a workaround to disable caps lock state. We probably need to add a request to the protocol when a virtual modifier is added.

I think we can merge this now, with the understanding that it will be improved later, and won't cause issues when the protocol isn't in use.

[Drakulix]

can we downgrade this to at least info (probably debug)?

[ids1024]

Oh right, that's a debugging line I added at some point that I didn't mean to commit. (Downgrading to debug would be appropriate).

[Drakulix]

Is this todo problematic? Can this currently interact weirdly with modifier only keybindings? (mostly for future reference to understand what needs to be done here.)

[ids1024]

I think that was just a comment to remind myself to look into whether this interacts problematically with that at some point. Not for any known issue.

I'll just remove that comment, but more testing could definitely be done in general.