Check dt in updateFromVelocity

[tonynajjar]

updateFromVelocity is not only called from update but also directly if position_feedback is false so we should also check there is dt is too small

[christophfroehlich]

I'm fine with this in general, but in case dt=0 nothing bad happens anyways?
we should also rewrite the API here, see #1394 for a first draft

[tonynajjar]

but in case dt=0 nothing bad happens anyways?

if it's 0, then you get a division by 0 further down

[tonynajjar]

There is more to be debugged as discussed but this PR can already be merged I would say. It already fixes a major bug for me because if one dt=0 goes through, the accumulators always return nan afterwards

[christophfroehlich]

Sorry, I got confused because the arguments left/right_vel are no velocities but position delta values. If it would be velocities we would not need to divide by dt in this method, and integrating with a step size of zero would just do nothing.

We really should rewrite this API, but I understand your point and we can merge this as a temporary fix.