Use the numpy Generator API uniformly

[jack-pappas]

riptable has a number of uses of the older numpy RNG APIs, e.g. np.random.randint(). The "new" Generator-based API is cleaner and often offers improved performance; it also avoids maintaining a global random state -- instead default_rng allows individual PRNG instances to be initialized with a seed value if desired.

riptable should uniformly use the new APIs; every method that uses these APIs should also be modified to accept an optional seed parameter (see Dataset.sample() for an example).