Add "Education" to the intended audience classifiers, since we use the example of using it in the classroom in our package description.

Get rid of linebreaks that aren't structural, so that distutils doesn't come along and mess up the rst.

Fixes #586, the error was not due to the values being compared, but to the
comparison function itself.

See http://projects.scipy.org/numpy/ticket/1684 if you're curious.

tests.

tensor/test_basic:T_scalarfromtensor.test0 which was already a known failure

This should unbreak GPU shared variables, broken after [c4a975e14f7c].

Instead of suppressing calls to free after closing the cuda context, delay closing the cuda context until all calls to free have been made. Might not be *quite* as performant, but feels perhaps a little cleaner.

Track whether there is a gpu context active; if not, don't try to free memory. Note that trying to free the memory and then just not displaying the error doesn't work, since the act of trying to free the memory creates a new gpu context.