make the api of the cuda shared variable constructor similar to the tensor shared variable constructor.

Added checks around call to CopyFromCudaNdarray in setitem, and added Equal method for comparing CudaNdarrays in C code

Added a few things to support GpuJoin: a ZerosWithPattern method to cuda_ndarray, and a slight refactoring of Join(Op) to inherit from this class in GpuJoin. And added GpuJoin itself. And a few unit tests for it.

numpy define the fct maximum and mininum as doing elemwise max and min.
We have the theano.tensor.{largest, smallest} that did that but with more op.
That make the graph harder to read. We will probably map largest and smallest
to the new fct in the futur after time show they are good.

Added setitem for CudaNdarray based on suggestions from Fred. Can only assign another CudaNdarray, and no support for broadcast in assignment, for the moment.

fix syntax error of tensor.ARange.infer_shape found by Justin Brayer and fix behavior error found while making test for that case.

Added test for this. The new graph are ugly... 12 ops with dimshuffles and rebroardcast...

 Optimization removed due to a bug noticed in the tutorials .. I will be trying to fix this as soon as possible.

This allow all other optimization to work correctly. This allow the insertion of gemm into gpu code in the deep learning tutorial of the mlp to have a speed up of 3x.

updated test to check for this.