Several problems with the previous documentation:

* It said it's stable "within the same binary" but not stable "across multiple binary invocations". Those statements contradict each other.
* It said "in the same binary will not produce the same sequence of variates within the same binary", which is repetitive since it says "in the same binary" twice.
* The comments about the seed sequence were not all next to each other, they were split apart by a rand.req.urng comment.
* It implied that process stability is all you get, whereas generally you also get changelist stability.
* It said you can use a seed sequence with SharedBitGen, but you can't.
* It said there is more info in std_seed_seq.h , but there isn't.

PiperOrigin-RevId: 501389708
Change-Id: I5e3dbc3548cc051265b8d004191c23147eccecc3