If a CudaNdarray has the right shape to be used as storage for a
broadcastable Variable, its stride should be 0.

Closes gh-597.

C code cache version

Added tests to check that we free all intermediate results on the gpu.

Instructions for manual Win 64 installation

This is needed to be able to version the CURAND op and all other op that have a generic in its input.

This is needed to cache the c code.

Test fix

Make CLinker use preallocated output memory

Commented pycuda example that use the pycuda elemwise generator is it work only with old version of pycuda.

Test fix

Fix ops to accept any kind of preallocated output memory

some extra verbose prints in gpu conv code.

Cache

add the theano flags cmodule.warn_no_version to print warning when c_code_cache_version is not implemented.

Update key for the compilation cache, the generated code is different,
we should force recompilation.

Argsort numpy wrapping