Now that the initial prototype is done, we should work toward passing tests and keeping them green. The goal of this is not to start fuzzing IonMonkey, but to make sure we're catching our own regressions and to weed out any edge cases that might break our design. Some of our test breakage is us not supporting certain operations late in the compilation pipeline, for example, unimplemented LIR. We can either change these assertions to compilation failures, or implement the ops on a case-by-case basis. (We'll have to implement them eventually, of course, but I'd prefer to get as many out of the way to keep pushing the design.)