GpuCorrMM: Removed pad='auto', added pad='half' and pad='full', added shape management for gradients

Adds caffe's implementation of the full convolution to CorrMM; cleaning up and documenting the code on the way

Auto generate Op.{hash,eq,str}

Fixed the compilation errors in mod.cu

[ENH] OpenMPOp, doc and cleanup

Moved `#include <algorithm>` outside the `#if PY_MAJOR_VERSION >= 3` block.

The generated mod.cu contains a number of std::min and std::max. Including <algorithm> will remove the not-found errors from these two functions.

Fix errors in SparseBlockOuterSS c_code.

Mixed stuff: pep8, comments, doc, don't enable GpuConv*3D op by default at that module import.

[opt warning] fix gpu reduce opt warning.