Consider specializing seed_from_u64 for Pcg32

[vks]

The default implementation of seed_from_u64 already uses the same algorithm, which could result in unwanted correlations. We should consider to provide a different implementation to avoid this problem.

[dhardy]

I don't believe this is an issue since the XSH RR output function is not linear, and also not injective (output is half the size of input). Further, a Pcg32 PRNG seeded like this would have a stream set, which has a very low chance of being 11634580027462260723 (the increment used by seed_from_u64). Thus I don't think there could be any issues more significant than weaknesses in Pcg32 itself.

@imneme are you aware of any issues seeding a Pcg32 PRNG from another Pcg32 PRNG?