Revert adding unit to pi due to regressions

[HansOlsson]

As indicated in #4046 it we need to revert part of #4155 since it is causing too many regressions for a minor release.

With proposed unit-checking in Dymola this reduces the warnings from 2301 to 1224 (Most - but not all are unit-related.)

Obviously many of them could be corrected, but I don't see that we have the resources to focus on that for this minor release. The first models found were CompareTransformers, SeriesResonance in Modelica.Electrical.Analog.Examples.

[henrikt-ma]

I would expect there to be a small number of root causes, and that we would have time to correct these before the freeze at the start of the next year.

Edit: Based on the affected examples mentioned above, the fix would be a simple matter of introducing parameters with the proper units instead of the literals used in the binding equations involving pi. The newly introduced parameters can be made protected when there is a need to protect agains backwards incompatibility in the future, if Modelica one day would have a mechanism for attaching units to literals (for example, L = (0.1)["V/A"] / (2 * pi)). Alternatively, the newly introduced parameters could be made public, which would actually add value to the examples as it makes it more natural to change values after translation.

[HansOlsson]

I would strongly recommend to merge this, and then prioritize investigating and fixing the issues together with other items.

[henrikt-ma]

I suggest that we get started right away cleaning up the uses of pi, and then we can see how far we get before the planned freeze early next year.