This provides performant length-indexed vectors with a suite of rewrite rules implementing fusion of intermediate structures, so that expressions involving many operations can usually be compiled to a single pass of computing and writing each index of the final Vec.