Implement Kalinski Modular Inverse & Benchmarks.

- Added `Criterion` on dev-dependencies and a
benches config section on Cargo.toml

- Added `benchmarks` folder with `benches.rs` file.

This file contains the benchmarks implementation of
the two distinct modular_inverse implementations.

Results:
```
Modular Inverse/Kalinski Modular inverse
                        time:   [50.105 us 50.438 us 50.819 us]
                        change: [-44.740% -44.195% -43.664%] (p = 0.00 < 0.05)
                        Performance has improved.

Modular Inverse/Standard Mod Inv
                        time:   [82.301 us 82.430 us 82.580 us]
```

To run the benchmarks just run: `cargo bench`.

Added `pub` visibility mark to `mod_inverse()`
fn to be able to benchmark it.

All tests are passing and Kalinski Inverse
has been introduced on the code replacing
the last mod_inv implementation.

Bitshift to the left instead of multiply by 2
since performs ffaster than mul.

Implemented Kalinski Modular inverse algorithm.
See: https://www.researchgate.net/publication/3387259_Improved_Montgomery_modular_inverse_algorithm

Also implemented test for big and small Bignums.

One test is giving results over the Montgomery Domain, so
a little bit of research has to be done.