* modify document of 11 APIs
test=develop
test=document_fix

* fix dtype to data type and description of name parameter

The macro should be defined by compiler rather than by source.

test=develop

* Fix infer shape of crop op.
* Speed crop op.

There are mainly following fixes:

- take `DeviceContext` as the template parameter of math functors and OpKernel instead of `Place`
- remove `eigen_device` interface in base class  `DeviceContext`
- remove `GetEigenDevice` interface in `ExecutionContext` and base class `DeviceContext`
- remove unused `platform::EigenDeviceConverter`
- rename `REGISTER_OP_GPU_KERNEL` to `REGISTER_OP_CUDA_KERNEL`
- rename `USE_GPU_ONLY_OP` to `USE_CUDA_ONLY_OP`

2. Fix unitest