Implement the no-overshooting condition for iterative amplitude estimation

[Cryoris]

What should we add?

The current implementation of iterative amplitude estimation does not include the no-overshooting condition outlined in the paper. This condition can significantly improve the performance and is likely required to obtain the performance shown in the paper.

Adding this should be relatively straightforward and could be a nice starting project!