jl_rng_split: not an problem, but weird these are all even; regenerate

This isn't an issue at all, but I noticed that these additive constants
are all even, which feels weird, so I regenerated them. Since we're
changing the fork behavior here anyway, changing these isn't any extra
disruption.

Author: StefanKarpinski

src/task.c

@@ -1061,10 +1061,10 @@ void jl_rng_split(uint64_t dst[JL_RNG_SIZE], uint64_t src[JL_RNG_SIZE]) JL_NOTSA
     // high spectrum multiplier from https://arxiv.org/abs/2001.05304
 
     static const uint64_t a[4] = {
-        0xe5f8fa077b92a8a8, // random additive offsets...
-        0x7a0cd918958c124d,
-        0x86222f7d388588d4,
-        0xd30cbd35f2b64f52
+        0x214c146c88e47cb7, // random additive offsets...
+        0xa66d8cc21285aafa,
+        0x68c7ef2d7b1a54d4,
+        0xb053a7d7aa238c61
     };
     static const uint64_t m[4] = {
         0xaef17502108ef2d9, // standard PCG multiplier