Update RAPIDS accelerated UDF examples for static CUDA runtime about spark-rapids-examples

[LIN-Yu-Ting]

Agree with this issue. Imagine that your native code jar is built in CUDA 11.8 environment and you try to run this in CUDA 12.0 environment then you might not be able to do that.

[GaryShen2008]

It seems simply to change the "CUDA_STATIC_RUNTIME ON" so that the udf.so won't link to cuda runtime dynamically.
One question is that if we build in the docker with cuda11, but run with a spark-rapids plugin cuda12, will it be a problem?
If so, we should clarify in README to avoid such case.

[jlowe]

It seems simply to change the "CUDA_STATIC_RUNTIME ON"

Yes, I believe it's essentially that, although there's a couple of spots that need to be updated due to lack of variable reuse. I'll post a PR shortly.

if we build in the docker with cuda11, but run with a spark-rapids plugin cuda12, will it be a problem?

No. Because we're statically linking the CUDA runtime, only the driver is required. The driver is backwards compatible with older CUDA runtimes.