implement limbs-based opcode circuit: MUL/MULH/MULHS/MULHU

Refer limb MUL/MULH circuit from here [mul](https://github.com/openvm-org/openvm/tree/main/extensions/rv32im/circuit/src/mul/core.rs), [mulh](https://github.com/openvm-org/openvm/tree/main/extensions/rv32im/circuit/src/mulh/core.rs)

We test constraints with fields on Goldilocks + e2e pass
