Fix hall sensor velocity estimation

[ahughesattomra]

The hall sensor velocity estimation could end up returning a nonzero value if the timer "wrapped" between runs, even if the motor was actually stationary. If using the velocity regulator the motor may then fail to start.

To solve this:

• Switch to using unsigned types so that wrapping is defined.
• When timing out the pulse value, also reset it so that the velocity won't be briefly treated as valid again when the timer wraps.

Seen on a platform with 32 bit long, so running the motor and then waiting ~40 minutes is enough to observe this bug. Note that the pulse_diff variable is reset if the hall sensor direction changes, which may happen naturally when the motor stops, depending on what it is attached to.

[ahughesattomra]

Only tested on v2.3.1 on a non-Arduino platform.

The bug is real though, so it'd be nice if it was fixed upstream.

[runger1101001]

Thank you very much for contributing this. I can't easily test it at the moment, but we'll review it and then include it in an upcoming release.