Remove `None` as a scheduling option

[pavel-kirienko]

This special case doesn't look convincing. I understand that there is some value in being able to unschedule a callback but as far as I understand by now we haven't identified an explicit need for that, and it would be easy to retrofit -- if needed -- later. The troubling thing here is that an extension of any well-structured set of entities with a special case often tends to break the design in unexpected ways at some point.

But I couldn’t resist the temptation to put in a null reference, simply because it was so easy to implement.

_Originally posted by @pavel-kirienko in https://github.com/OpenCyphal-Garage/libcyphal/pull/371#discussion_r1704753104_

[serges147]

Notes:

• [x] Eliminate None schedule.
• [x] Move public Callback::Handle -> protected SingleThreadedExecutor::CallbackHanlde.
• [x] The case of callback schedule cancellation could be covered by adding (in the future, if needed) new callback.cancel() (or cancelSchedule | unschedule, TBD) method - callback will remain fully functional but just in a dormant/unscheduled state, ready to be resheduled.
• [x] executor.scheduleCallback(callback, ...) -> callback.schedule(...) move will eliminate another point of failure which is currently described in docs as "...must be used only with the same executor..."
• [x] An attempt to schedule already .reset() callback (the one which had released all the resources, function and reference to its executor) should assert. As an alternative to the assert, research possibility to have bool (non-discardable? TBD) return value from the .schedule method (probably not for the .cancel).

[serges147]

Closing as fixed by PR #376