randr: Add mirroring support

[Drakulix]

This does a bunch of refactoring to cosmic-randr to add support for our upcoming cosmic-output-management protocol: https://github.com/pop-os/cosmic-protocols/pull/26.

Most notably, it cleans up a bit of the wayland-rs usage:

• Gets rid of useless Data-type
• Uses more userdata instead of ObjectId-hashmaps, where it makes sense (e.g. for Modes)

Additionally

• Store scale in a more precise f64 (e.g. 130% can't be cleanly represented without more precision)
• Bind cosmic-output-management when available and use it for more precise scale and mirroring.
• Actually always send enable/disable/mirror requests for all heads as required by the protocol
• Automatically perform the initial dispatches on connect to populate all data fields
• Don't bind whatever version of wlr-output-management the compositor advertises, but the highest known one (4 in our case).
• Only send release-events for supported versions

Missing features:

• [x] KDL output isn't touched, as I am not sure what cosmic-settings expects and didn't want to introduce accidental breakage
• [ ] No way to set adaptive-sync (although it is reported now). Currently this feature can still be quite buggy, so this is left to be enabled later, once necessary refactoring/bugfixing in cosmic-comp is done

Should be tested against https://github.com/pop-os/cosmic-comp/pull/448

[mmstick]

cosmic-settings uses the shell crate here for getting information from cosmic-randr, so you're free to modify the KDL output as long as the shell crate can still deserialize it.