[Bugfix] GptJ & GptNeoX batch inference error

[YZP17121579]

GptJ & GptNeoX may generate random outputs when using batch inference mode and no prefix prompt.
The problem is caused by the nullptr check
in

FasterTransformer/src/fastertransformer/kernels/gpt_kernels.cu

Line 1064 in f8e42aa

if (tiled_prefix_prompt_lengths != nullptr) {

[BasicCoder]

I think this is a duplicate solution of #716 which is more elegant and efficient.