Parallelize the flat Karatsuba implementation

[emmorais]

Parallelize flat Karatsuba:

• [ ] Each layer of the flat Karatsuba computation has independent chunks that can be calculated completely in parallel. The results must be ready for next layer computation.
• [ ] When we get too close to the top of the calculation tree, parallelization will be less efficient, then we need to fine tune a parameter that will dictate when to stop using parallelization.

[teor2345]

I've marked this as wontfix, because:

• The recursive implementation is already 4-5x faster: https://github.com/Inversed-Tech/eyelid/pull/32#issuecomment-2054782381
• Parallelising computations that take a few milliseconds usually doesn't provide much performance gain, because of thread overhead

Instead, we could parallelise multiple separate iris comparisons or polynomial multiplications.

[teor2345]

We’re more likely to parallelise the overall comparison, rather than individual operations. It’s more efficient to do that.