In order to avoid expanding memory usage and computations in the part of
the graph that computes gradients, I propose the following conventions,
that re-instate some of the constraint that existed before on the
dtype of gradients:
- When calling some_op.grad(inputs, output_grads), each variable in the
  "output_grads" list, if it is an actual numeric variable (and not,
  for instance, DisconnectedType or NullType), should have the same
  dtype as the corresponding output variable.
- Moreover, if one of the output variables is of a discrete dtype (int
  or uint), then the corresponding output gradient (if not a special
  case like NullType) should be zeros.

This is implemented in theano.grad, so the Op's grad method does not
have to be changed, but now it can rely again on the fact that, if an
output gradient has a dtype, that dtype will be the same as the
corresponding output variable.

Fix clip grad

MPI Ops

One of the mpi tests needs to run a file in the theano file system.

It is now pointed to that file through the theano.__file__ attribute.

mpi_cmp -> mpi_cmps

fix incorrect test `y = x + x` -> `y = x + 1` to match expected output

remove old comments

(these functions moved in a rebase)

I'm currently using os.popen.
I tried using subprocess.Popen without success. Would appreciate tips.

Also removed infer_shape for now

I suspect that this is necessary to hold as an output

Revert "Merge pull request #995 from pascanur/fix_grad_scan_dtype"

This reverts commit 4ecb5a04, reversing
changes made to 8b088c69.

Fix grad scan dtype

Remove debug line following code review.