Documentation mistake for SegSortPairsFromFlags

[Yuduo Wu]

On the documentation for SegSortPairsFromFlags, the sequence of parameters (count and flags_global) seems disordered.

Should be(KeyType* keys_global,
ValType* values_global, int count, const uint* flags_global,
CudaContext& context, Comp comp, bool verbose = false) instead.

However, the following is listed on the documentation currently:
// Segmented sort using head flags and supporting value exchange.
MGPU_HOST void SegSortPairsFromFlags(KeyType* keys_global,
ValType* values_global, const uint* flags_global, int count,
CudaContext& context, Comp comp, bool verbose = false);

Thank you!