Fix BigInt Type Inference

[tzerrell]

Make a number of fixes to BigInt dialect type inference. The bulk of the work is in https://github.com/risc0/zirgen/pull/45. What's in this PR is largely the change to the allocation in risc0/circuit/bigint/src/byte_poly.rs (since we were allocating one more coefficient than needed during polynomial multiplication) and switching RSA tests over to use the test harness. The other changes (to control IDs and tests) can be generally described as internal changes that needed to happen as consequences of the regenerated BigInt code.