2. Make SwitchOrderLayer support for softmax activation
3. Fix bugs

Tensor should use PADDLE_ONLY_CPU macro

Add voc2012 dataset for image segment

Enable error clipping for fc layer

fix travis check style for Go

Override Net::DebugString to invoke inner ops_

Add method converting Tensor to Eigen TensorMap

Make CreateOp in Plain C++ params

Read/Write a Tensor Python

* Not use __CUDACC__ macro, because malloc in GPU memory do not have to
  used in .cu file.
* Also, `PADDLE_ENFORCE` logic is wrong.

Change `in_out_idxs_` to shared_ptr

Add gometalinter errcheck to pre-commit-hook, fix all Go unchecked errors.

* `in_out_idxs_` shares between all operator instance in same type
  of operator.

Fix Merge Bugs

Net::DebugString() print what it is and invokes inner operators
DebugString() with a prefix '\t'.

* There is a merge conflict when merge PR #2914
* Develop and PR #2914 both add `DDim::size` method, but did not
  triger git merge conflict while merge.

Add skeletons of `mul`, `rowwise_add`, `sigmoid`, `softmax` ops

* Follow review comments to seperate Tensor Numpy interactive methods in
  tensor.h.
* Simplify logic for `CastToPyBufferImpl`, make it as one struct and in
  details namespace.
* Remove `Scope` expose in Python, since it currently is useless.
* Remove some debug functions.

Fix some link errors about NNPACK.

refine tensor's conditional compilation, remove Tensor::numel_ and add DDim::size()

Unchecked errors could be handled by: cd go; gometalinter --vendor --disable-all --enable errcheck $(glide nv)

Basically following
http://pybind11.readthedocs.io/en/stable/advanced/pycpp/numpy.html

* Use buffer protocol to return a view of Tensor. It can be cast to
  numpy array in Python.
* Set a numpy array to a tensor.