feature: seeded random sampler

[FelixAbrahamsson]

Alternatively we create a new sampler?

[samedii]

I guess this could be useful to have still and I was going to approve but it seems that the generator is changing the global random state which is probably not what we want? I am also not sure if the sampler is lazy when sampling or eager. We might want a for-loop that yields samples and switches the random state

This is still equivalent to a SequentialSampler on a shuffled dataset right so maybe there is a simpler alternative?

[samedii]

Do we still think this is useful? Have you tried it out more @FelixAbrahamsson ?