Cuda fortran order

Alloc cst fold

Remove condition that is always True as we check if it was "output" before and it is the only string supported.

Since we do constant folding earlier, this make a GpuAlloc test fail.
This could also speed some code as this allow more op to work inplace.

Faster opt

I review the code, we always put a tuple or a Theano variable in that
list. We only want to work with the tuple case. So just test for the
tuple. We don't want to cover the case where we have a Theano vector
with a know lenght.

also a few pep8, as we support python 2.6+ now this is valid.

This give a small speed up.

Also import a module in an opt only once.

This make the graph smaller faster, so it speed up optimization

Exclude -avx as a valid extention for gcc 4.7.3 due to a bug report from...

Add python platform-dependent C include files to the include directories

pyconfig.h maybe located in the platform-dependent include directory

Add a_tensor_variable.cumprod/cumsum() interface 

Fix cuda test on 32 bit python.

Softmax optimization

Make use of the verbosity parameter in numpy system_info.

is avaiable not just in the inner of the loop for the Cross entropy
fct. This also remove duplicate code.

bugfix, for a case that don't happen right now.