This  make  the  graph  before  compilation  and  when  compiled  with
FAST_COMPILE   more    readable.   This   was    optimized   away   to
local_fill_to_alloc, but if  we enable it in fast_compile,  it also do
more optimisation that make the  graph not readable without enabling even
more optimization in fast_compile.s

Fix test

Fix scan debugmode

Fix scan debugmode

destroy_map is not always defined for a scan op.

Fix tests

Looks ok to me.

Pycuda init

This is probably due to wrong handling of context. I think theano and
pycuda don't use the same context while the should. I am investigating
the problem, but while this make the buildbot segfault, I disable it
for now.

Update to sparse sandbox

better error msg.

Mixed