Add support for seeding model

[rfriedman22]

Currently there is no support for seeding random number generators, which would be useful for reproducibility purposes. A user can seed the RNG on their own using the following three lines:

random.seed(seed)
np.random.seed(seed)
tf.random.set_seed(seed)

However, it might be helpful to build this into model.fit by providing the seed as an optional argument to the method.

[atareen]

Hi Ryan,

Thanks for writing. We tried implementing what you suggested in Mave-nn prior to the release of version 1.0.1, nearly exactly as you have suggested. With the version of Tensorflow we were using at the time (close to 2.0.0) we found that models still didn't behave identically, even after specifying all these seeds.

Mave-nn is undergoing active development, to be released as Mave-nn2 some time in the near future, and I can re-try to fix the seed so that Mave-nn2 gives identical models.

Thanks,
Ammar