I've stripped out everything not strictly required for mouse button HID
support. This include dropping the worker thread completely, shaving
down the USB HID descriptor. Flattening the mouse/ source directory
structure and removing a bunch of event handling.
I have kept the mouse event handling separate from the other HID event
handling since I figured that was a pretty neat split. If that's a bad
idea, do tell.
I've also added a test case for mouse button emulation, since that was
untested before.
The changes have been tested on a corne (split) in usb mode. Bindings on
both the left and the right side works (with the left side as master).
Added a new setting to remember the user's preferred endpoint. When both USB and
BLE are connected, the preferred endpoint will be used.
Added a new behavior to control this setting. It supports commands:
&end END_USB - Prefer USB output
&end END_BLE - Prefer BLE output
&end END_TOG - Toggle between USB and BLE
* Add `bt` behavior that can be used to perform certain actions,
such as next/prev identity, reset identity, etc.
NOTE: Multiple identities is only supported for non-split shields,
due to missing Zephyr identity functionality for dual
central/peripheral devices.
* Proper bond reset tied to action, that honors peripheral bonds,
so folks can reset and pair to other hosts, without breaking
bonds between splt halves.
* Add initial event manager implementation,
roughly mimicking Nordic's API.
* Add `position_state_changed` and
`keycode_state_changed` events.
* Hook up HID and keymap to new events
instead of using behaviour global event
crazy.
