Implement "optimizer" versions of the rise/set/transit functions

This is an idea that struck me a couple different times but I'm not sure I've ever written it down, but it has now been re-prompted by #424 and #440... Should we have an option for the rise/set/transit functions to use an optimizer instead of a grid-based approach? The grid is definitely better for large numbers of targets or many nights, but for just one-night, using an optimizer from the outside to find the zero-crossing might be faster/more precise?

(Caveat: as @bmorris3 mentioned in https://github.com/astropy/astroplan/pull/424#issuecomment-543173951 it's not clear that "more precise" is that critical since most of the time these approaches are not particularly *accurate* without a lot of knowledge of atmospheric conditions... But that doesn't mean we shouldn't allow the *option* since someone might actually have really good atmospheric data some of the time...)

Start by reviewing the rise, set, and transit functions and the context in issues #424 and #440. Determine whether an optimizer option is wanted, how it should coexist with the grid-based approach, and what precision and performance tests would define a completed implementation.

Written by the indexing model from the issue text.