chore: mitigate copies in final_exponentiation

[arthurpaulino]

This PR applies the techniques from #11 in order to further reduce copying in the hot path of pairing.

[arthurpaulino]

@wwared I've deleted Fp6::mul_inp and instead did a small optimization in Fp6::mul_interleaved that resulted in a cheap drop of 100k cycles

[huitseeker]

This looks great, but it includes a lot of APIs that are essentially one & away from the original. I wonder: how hard would it be to write tests for those functions comparing them to the original?

IOW:

const TEST_CASES: usize = 100;

#[test]
fn test_foo() {
  for _i in 0..TEST_CASES {
    let a = Input::random();
    let b = Input::random();
    assert_eq!(foo(a.clone(), b), foo_inp(&mut a, &b));
  }
}

?

[wwared]

@huitseeker: Opened issue #13 to track this, I think it shouldn't be too hard (requires to be careful around the functions that got cfg'd out of the zkvm), but we definitely want to do this