Allows unpickling gpu cudaNdarray as numpy.ndarray.
See AddConfigVar comment for more details.

There is no more use for choosing single tests to skip.

This reverts commit 99295012.

However, emit a warning.

Conversion of GpuSoftmax and GpuSoftmaxWithBias to the new backend

Fix unpickling of Elemwise op since openmp changes.

Tag slow tests

Don't try to use the c_code for GpuElemwise on opencl since it's cuda-only.

Replace content of c_init_code by reference to function setup_ext_cuda() in GpuSoftmax and GpuSoftmaxWithBias

Elemwise openmp

Fix Sparse grad crash and implement mixed dtype in sparse Mul/Add

I changed how format is specified. There was no bug as the input
variable are put in the key. Now I make explicit in the op key the
format.

This simplify at a few places the code too.