Refine operator cmake

[typhoonzero]

• move operators of same kind to several folders
• simplify operators/CMakeLists.txt

simplified operators/CMakeLists.txt:

include(operators)

# clean cache and pybind_file content first when rebuild
unset(GLOB_OP_LIB CACHE)
unset(OP_LIBRARY CACHE)
set(pybind_file ${PADDLE_BINARY_DIR}/paddle/fluid/pybind/pybind.h CACHE INTERNAL "pybind.h file")
file(WRITE ${pybind_file} "// Generated by the paddle/fluid/operator/CMakeLists.txt.  DO NOT EDIT!\n\n")

add_subdirectory(math)
add_subdirectory(controlflow)
add_subdirectory(csp)
add_subdirectory(detection)
add_subdirectory(elementwise)
add_subdirectory(fused)
add_subdirectory(metrics)
add_subdirectory(optimizers)
add_subdirectory(reduce_ops)
add_subdirectory(sequence_ops)

if(WITH_DISTRIBUTE)
    add_subdirectory(distributed)
    add_subdirectory(distributed_ops)
endif()

if (NOT WIN32)
    add_subdirectory(reader)
endif()

if (NOT WIN32)
    add_subdirectory(nccl)
endif()

if (WITH_GPU AND TENSORRT_FOUND)
    add_subdirectory(tensorrt)
endif()

register_operators()

set(COMMON_OP_DEPS "")

set(COMMON_OP_DEPS ${COMMON_OP_DEPS} xxhash selected_rows_functor selected_rows lod_tensor maxouting unpooling pooling lod_rank_table context_project sequence_pooling executor dynload_warpctc sequence_padding sequence_scale cos_sim_functor memory jit_kernel concat_and_split cross_entropy softmax vol2col im2col)
if (NOT WIN32)
  set(COMMON_OP_DEPS ${COMMON_OP_DEPS} sequence2batch lstm_compute matrix_bit_code gru_compute activation_functions)
endif()
if (WITH_GPU)
  set(COMMON_OP_DEPS ${COMMON_OP_DEPS} depthwise_conv cub)
endif()

[luotao1]

LGTM!!