I threw this together in a day, somewhat based on my Conway's Life code. It's set to use MMC1 but doesn't actually write to any MMC1 registers, only using it for the extra RAM emulators grant it.
This demo runs breakout, using left and right to move the paddle. You could replace it with something else, though Blinky wouldn't run when I tried.
One big problem I ran into is that despite the sprite drawing being incredibly inefficient (calls a routine for every single pixel that should be flipped), games still run way too fast. If games actually used the timer it would be fine, but many don't seem to, so I need some sort of instruction limiter. This ROM just runs one instruction per frame.