Improvement on the integral contribution implementation

[carlosjoserg]

The integral of the error is computed in a class member, and the integral contribution is computed as of it, instead of doing all together. Now, setting the integral gain to zero dynamically cancel the integral contribution.

Additional update:
Function getCurrentPIDErrors(...) now returns the integral of the position error, and not the contribution. Because of this, also the tests were modified. Tests integrationWindupZeroGainTest (actually, this one makes no sense now, since the error is directly given instead of obtaining by diving the term by the gain) and integrationWindupTest check the windup constraint on the computed command.

I don't know how this affects the dependant packages, but I believe the change is worthy.

Resolve #32.

[adolfo-rt]

LGTM.

On getCurrentPIDErrors(...) returning the integral error (correct since this patch) vs the integral term (incorrect, existing behavior), I don't know if the existing broken behavior is used by anyone.

I'm going to let this one soak another day to see if somebody wants to comment on this particular issue, otherwise I'll give it a go.

[adolfo-rt]

Thanks for the two fixes!.