Ensure all usage of `ndim` to be 0-d tensor compatible

[junrushao]

EDIT: See also #14323.

After #14315, we should check all the occurrence of ndim to ensure they are used correctly in the semantics of 0-d tensor.

@reminisce @szha @eric-haibin-lin @zheng-da @yzhliu

List of occurrence

src/c_api

• src/c_api/c_api.cc
• src/c_api/c_api_common.h
• src/c_api/c_api_symbolic.cc
• src/c_api/c_predict_api.cc

src/common

• src/common/serialization.h
• src/common/static_array.h
• src/common/utils.h

src/executor

• src/executor/infer_graph_attr_pass.cc

src/imperative

• src/imperative/cached_op.cc
• src/imperative/imperative_utils.h

src/io

• src/io/image_aug_default.cc
• src/io/image_det_aug_default.cc
• src/io/image_io.cc
• src/io/image_iter_common.h
• src/io/iter_batchloader.h
• src/io/iter_image_det_recordio.cc
• src/io/iter_image_recordio_2.cc
• src/io/iter_libsvm.cc
• src/io/iter_sparse_batchloader.h

src/kvstore

• src/kvstore/gradient_compression.cc
• src/kvstore/kvstore_dist.h
• src/kvstore/kvstore_dist_server.h

src/ndarray

• src/ndarray/ndarray.cc
• src/ndarray/ndarray_function.cc
• src/ndarray/ndarray_function.cu
• src/ndarray/ndarray_function.h

src/nnvm

Nothing here.

include

• include/mxnet/ndarray.h
• include/mxnet/operator.h
• include/mxnet/operator_util.h
• include/mxnet/resource.h
• include/mxnet/tensor_blob.h
• include/mxnet/tuple.h

[mxnet-label-bot]

Hey, this is the MXNet Label Bot.
Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it.
Here are my recommended labels: Build

[reminisce]

Also include all the infer shape functions of operators.
#14323

[junrushao]

@reminisce Yep. I intentionally ignored src/operator in this list because there are too many occurrences in that folder, and we should share efforts across these two things - logic and implementation

[piyushghai]

@mxnet-label-bot Add [Numpy]

[piyushghai]

@mxnet-label-bot Update [Operator]

[junrushao]

Going to act on this thread this week. I guess most of them have been addressed in recent PRs like #14361, #14315, #14409.

[junrushao]

#14612