Fix position prescribed friction

[m-kormann]

The fix from #4129 has been accidentally reverted by #4255 without being the reason for the compiling errors.

The compilation errors should be fixed by recompilation of the binaries as discussed in #3911, #4178 and #4250.

[DagBruck]

An improvement compared to the status-quo. As discussed in the earlier PR 4129.

[MartinOtter]

I have re-analyzed the solution and propose a slightly different fix. I would have included this fix in the pull request, but I don't manage to quickly do it (and do not have enough time to figure this out). So, someone else has to include this solution.

Issue with current proposal:

a_relfric/unitAngularAcceleration = if locked then 0 else if free then sa
else if startForward then sa - tau0_max/unitTorque
else if startBackward then sa + tau0_max/unitTorque
else sa - sign(w_relfric)*tau0_max/unitTorque;

Assume mode == Forward, so forward sliding, and wrelfric crosses zero. Then previously, the equation "mode = .... w_relfric > 0 ..." triggered a state event at w_relfric=0 and during finding the root w_relfric=0, all signals are continuous. With the change above, this is no longer the case, because sa and a_relfric are computed with "a_relfric = sa - sign(w_relfric)..." during root finding, so a_relfric and sa are discontinuous during root finding (which is bad).

This can be fixed by reformulating the equations to:
a_relfric/unitAngularAcceleration = if locked then 0 else if free then sa
else if startForward then sa - tau0_max/unitTorque
else if startBackward then sa + tau0_max/unitTorque
else if pre(mode) == Forward then sa - tau0_max/unitTorque
else if pre(mode) == Backward then sa + tau0_max/unitTorque
else sa - sign(w_relfric)*tau0_max/unitTorque;

The difference can be seen in the test example ModelicaTest.Rotational.TestFrictionPosition with a small number of output points ("number of intervals = 10" in Dymola).

Proposed solution

New solution:

[tobolar]

I have re-analyzed the solution and propose a slightly different fix.

@MartinOtter I have implemented your suggestion in PR m-kormann#1
@m-kormann Please merge m-kormann#1 on your branch "revert-4255-revert-fix-friction" in order to be visible here.