Make `EValueEstimator` accept a `Generator`

Currently the only way to ensure reproducible results of the `EValueEstimator` is calling `np.random.seed()` before creating it. However, this `np.random.xxx` API is deprecated and superseded by using an explicit [`Generator`](https://numpy.org/doc/2.0/reference/random/generator.html).

To reflect this change in `numpy`, `EValueEstimator` should accept a `Generator` to ensure reproducibility.