Fix x11 DeviceEvent ModifiersStates

[rukai]

The other half of the fix for #226
I put this into a seperate PR because it will likely be held up by discussion.

Ralith said on IRC: "looks like I just didn't implement modifiers for raw kb input"
However looking into it I couldnt find a way to do so.
(Because the XIRawEvent doesnt have the state field that XKeyEvent does)

Because DeviceEvents dont require ModifierStates like WindowEvents do, as ALL events are received regardless of focus, I decided to create a seperate DeviceKeyboardInput without ModifierStates.

There are currently no other backends returning KeyboardInputs in DeviceEvent's so I couldnt check how they handled it.

Interested to hear your thoughts @tomaka @Ralith

[Ralith]

Under XInput2, you're supposed to track keyboard modifier states yourself using libxkbcommon, specifically the xkb_state_mod_* functions. Unfortunately there doesn't seem to yet exist dynamically-loading libxkbcommon and libxkbcommon-x11 bindings for rust, but it's not a large library, especially since we only need a small number of functions (xkb_x11_setup_xkb_extension, xkb_x11_keymap_new_from_device, xkb_x11_state_new_from_device, xkb_state_update_mask, xkb_state_key_get_one_sym, and xkb_state_key_get_utf8) used in the documented pattern.

I believe pulling in xkbcommon will also allow us to drastically improve the X11 backend's VirtualKeyCode support, by maintaining a parallel keyboard state (per device) that ignores modifiers, providing a straightforward source for stable keysyms we can map onto known virtual keys.

[Ralith]

On further research, it looks like an easier workaround might be XkbGetState, which is bound by x11-dl already. The downside is this round-trips to the X server, which isn't great.

[rukai]

Thanks for your research Ralith!
This looks more complex then I would like to get into, so I'll leave this for someone else to do.