This is to allow for LEDs controllable via transports such USB and I2C
where waiting (sleeping) is expected.
Previously the calls were made with a mutex held or even from a callout
The easing of restrictions comes at a cost of having a dedicated kernel
thread for LED blinking duties.
I have not decided yet whether it's okay to "permanently" set a LED via
a direct call or if it would be better to defer the setting to the thread
as well. The former would have an advantage of decoupling the LED control
logic from the caller's context. The disadvantage is obvious, the
operation would become asynchronous.