Fix handling of != and == in event triggers

[dweindl]

The recently introduced handling of == and != (#2708, #2716), while logically correct, does not work in the context of root finding.

For example, a != 1 will results in r(t, ...) = Max(1 - a, a - 1) which is 0 for a = 1, so an event will trigger although it should not. The same is true for any "hidden" !=, as in a > 1 OR a < 1 (which is what is currently done behind the scenes).

Handling of == and != should be restricted to fixed-parameter arguments (useful for indicator variables) and implemented as boolean expression in that case.

[FFroehlich]

This is already mostly covered by #2840

[dweindl]

This is already mostly covered by #2840

Thanks, will have a look.