Handle attached-detached devices with udev monitor

[kikadf]

(note: avoid the merge conflict it is based on #6)

With this change, libopeninput uses the udev monitor to detect device attach and detach events. The udev monitor only detects changes; wscons_udev_handler() calls libinput_path_add_device() or libinput_path_remove_device() to update the device list, and post_device_event() to send events, similar to libinput_udev_assign_seat().

interface_backends now only track the udev monitor state; all device handling occurs in wscons_udev_handler().

Tested on NetBSD with labwc: devices work correctly after detach/attach, and multiple mice are handled properly.

To test on OpenBSD, add a small patch to libudev-openbsd to handle wskbd and wsmouse devices, e.g., add the following to subsystem_config subsystems[] in udev-utils.c:

{
	.subsystem = "input",
	.syspath = DEV_PATH_ROOT "/wskbd[0-9]*",
	.create_handler = create_keyboard_handler,
}, {
	.subsystem = "input",
	.syspath = DEV_PATH_ROOT "/wsmouse[0-9]*",
	.create_handler = create_mouse_handler,
}, 

[mherrb]

On OpenBSD this doesn't work because 1) libudev-openbsd doesn't trigger on wskbd/wsmouse plug/unplug events (it's more or less restricted to USB fido keys) 2) when patched to handle wskbd/wsmouse, there is the ownership issue of the /dev/ entries (and and adding /dev/ws{kbd,mouse}[0-9]* to /etc/fbtab is not a option.
It doesn't seem to make things worse though.
Keeping that open for now.

[kikadf]

Thanks for your review. I tried to add wskbd and wsmouse support to libudev-openbsd, and you are right, it is not as simple as I initially thought.