[WIP] Don't add inequalities when not necessary

[Robbybp]

There are two situations where adding inequalities when eliminating bounded variables isn't necessary:

1. The bounds are dominated by bounds on the "defining expression". This can be determined by propagating bounds from the "independent variables" to the "defining expression" with FBBT.
2. The defining expression is linear (monotonic is sufficient) and contains a single variable. In this case we just propagate the bounds from the eliminated variable to the variable used to eliminate it.

I'm testing out these ideas. If they work well, I may make them the default. Otherwise I'll add them as options (or just for get about them altogether). I'll post results in #15. Opening this PR for reference.