New example of custom operator using RTC

[asitstands]

Description

example/numpy-ops/ndarray_softmax.py contains a custom softmax implementation using RTC. However, it extends NDArrayOp which is deprecated. Moreover the example seems not compatible with the current RTC API. I think that the example confuses newbies like me who didn't know the previous history of the API changes. So I think it is a good idea to remove the broken example and add a new RTC example based on CustomOp. The new example example/numpy-ops/custom_softmax_rtc.py is based on the existing example (example/numpy-ops/custom_softmax.py) but uses custom CUDA kernels through RTC.

Checklist

Essentials

• Passed code style checking (make lint)
• Changes are complete (i.e. I finished coding on this PR)
• All changes have test coverage:
• Unit tests are added for small changes to verify correctness (e.g. adding a new operator)
• Nightly tests are added for complicated/long-running ones (e.g. changing distributed kvstore)
• Build tests will be added for build configuration changes (e.g. adding a new build option with NCCL)
• Code is well-documented:
• For user-facing API changes, API doc string has been updated.
• For new C++ functions in header files, their functionalities and arguments are documented.
• For new examples, README.md is added to explain the what the example does, the source of the dataset, expected performance on test set and reference to the original paper if applicable
• To the my best knowledge, examples are either not affected by this change, or have been fixed to be compatible with this change

Changes

• example/numpy-ops/ndarray_softmax.py is removed.
• example/numpy-ops/custom_softmax_rtc.py is added.
• example/numpy-ops/README.md is updated.

[piiswrong]

great example. Thanks!