We don't want scalars to be converted into TensorScalars ( the pushout
optimization will fail if that happens)

Those used to belong to the original implementation of the optimization,
which I changed to fit the new equal_computation function.

It used to be used by the old implementation of equal_computations, but now
is not needed anymore

There is no reason for this, just that I had a lot of problems with this
optimization. It turned out the problem to be caused by
local_useless_subtensor optimization (fix which I will push later).
In my attempts to debug it though, I ended up converting it into a global
optimizer, and I don't want to go through the pain of reverting it and
making sure also that it works.

Basically add information about scan if it stops on a condition or not
(using keyword for or while)

Fix following Pascal suggestion of the __eq__ of scan.

It helps by compiling the inner graph only once, making compilation much
faster

Basically change the code to be aware that the last input can be a
condition, and if so stop when that condition becomes True.

Specifically for initial states, they should always stay variables, and not
be constants !

My optimization pushes in the main graph a lot of computations in terms of
constants, and constant folding maxes out pretty quickly. This is a quick
fix to that. I'm not sure if 50 is not too much, I haven't played with that
value .. maybe 20 or something would suffice. I would also like to increase
this value only for constant folding and not all the others, but I didn't
know how.

theano.until

(note : the implementation of make_thunk is seen by DebugMode as a python
implementation, and it sees no C implementation for that Op. Are there
checks for the C implementation that are not done for the python
implementaion ? )

treat them like any other theano tensor)

Note: this optimization expects the state of scan to have the as_while flag
and the profile flag.