Extend variability rules and fix invalid example

[henrikt-ma]

Fixes #2525

[HansOlsson]

Looks good, but will wait a bit for other feedback before merging.

[gkurzbach]

Maybe I missed some discussion but
b2 = noEvent(x > 1);
for me is clearly a variability error, because the right side may change outside of events but b2 not.

and a further question is whether the assignment from i2 = x; makes x to a discrete variable, because it cannot change between events.

[henrikt-ma]

@gkurzbach, I think we all agree that there is a problem with the equation, the question is how to capture this intuition in the specification. the reason we can't fail the normal equation b2 = noEvent(x > 1) due to variability error is that a normal equation has no prescribed direction of causality. This is different in algorithms, inside when-clauses, and for declaration equations. The point of this PR is to improve the variability rules so that this equation can be reported as an error with reference to variability rules in the specification, instead of needing to come up with a tool-specific explanation.

[henrikt-ma]

Regarding i2 = x, and as seen in the ongoing discussion in #2521, I think there is disagreement regarding whether x should be considered discrete-time or not for purposes of variability analysis.

This is my personal opinion on the matter: We shouldn't rely on this type of inferred variability when defining the variability of a variable. Instead, I'd like to be able to declare x as discrete Real, and that i2 = x; would be a valid equation for x (that is, not requiring x to be assigned in a when-clause).

For now, I suggest that x is just considered a non-discrete-time variable for purposes of variability analysis, that is, that we are not allowed to make use of the obvious fact that x has a discrete-time solution.

[gkurzbach]

To the first point: I got it now. To the second: I agree with you.

[gkurzbach]

So it seems to be ok.