Implement analytic integration for 1D models

[pllim]

Unlikely #88 would happen anytime soon, so just go ahead an implement it here for the models that we use. No point to use trapezoid integration for simple things like Box1D.

[pllim]

astropy/astropy#5033 likely not going to happen either, so we're back to square one.

[pllim]

Maybe I should just do something here without worrying about moving the code upstream to astropy first. Cases for 1D should be simple but not so for 2D (as evident in the discussions over at astropy/astropy#5033), but we only care about 1D here.