intrinsic-test: combine C files for faster compilation

[folkertdev]

There are 3 main changes

• emit the arguments (the arrays of inputs) inside of the test function, and rename the test function from main to run_{intrinsic_name}.
• generate number-of-cores files, and evenly divide all intrinsics over these files. Each file now has the functions for running many different intrinsic tests
• generate a main that is just a gigantic switch to select the right test to run

Parts of this are still inefficient, and for the full effect the rust side also needs work, but I suspect this is the biggest win (if not, it's actually cargo, we'll see I guess), and sort of a manageable change to review (hopefully).