This allow theano.tests() to run more tests!

This change the behavior a little. Now int are casted into
numpy.int64. This is needed as in Python3 long have replaced int and
there are no more long! So the int case will be used in Python 3.

Forcing the type isn't a problem as we support only this type for
shape everywhere in Theano.

FIX: Do not explicitly link against libpython2.x on OS X.

Updated custom BLAS compilation doc for Windows

Now describing how to compile OpenBLAS rather than GotoBLAS.

Minor fixes

Elemwise faster when c or f contig inputs

This speed up hard_sigmoid and ultra_fast_sigmoid by 2x.

We ignore input that are broadcasted scalar when checking for fortran
inputs.

Do not work yet when the inputs is fortran, as the outputs is always c contig for now.

Add hard_sigmoid, faster, but more approx.

Add ultra_fast_sigmoid

Collections.Counter added for python < 2.7