This stability optimization didn't work for at least three reasons:
1. It used tensor.neg(x) to identify -x, while in canonicalized form -x
   is written -1 * x.
2. There was a typo (calling a function sigm that didn't exist).
3. It wasn't able to parse a tree of multiplications, for instance
   something like (exp(x) * 2) * sigmoid(-x).

This new version fixes all three issues. In order to solve problem #3,
it uses a different approach that consists in parsing the optimized node
into a so-called "multiplication tree".

Currently, the structure of this tree is left unchanged. It would be
possible to have this optimization also flatten the tree, which would
make the code simpler (but potentially undo work performed by some other
optimizations).

Make scan opt before blas opt.

Fix neibs

Spectral radius bound review

Depricated scan

When tring to make the grad raise an Exception with an error that the old version returned wrong answer.

Ensure sorted indices reviewed

Fix default axis for tensor.roll

In the case of a default axis argument and a 1-d tensor, don't create
flatten and reshape nodes that need to be optimized out when it's simple
enough not to do so.

NumPy docstring standard compliant. Nearly identical to the numpy.roll
docstring.

Elegantly solved with a recursive call.

added roll function and test

small fix to gpusum. It can happen that we start 2 reduction at the same time

added roll import to tests

fixed pep8 issues

Better pushout optimization

Note that I only went through the files and fix whatever the PEP8 script for
vim reported as incorrect, but I had not actually paid attention to the
text, so there might still be inconsistencies (like for example strange
variable names, etc.). Also only scan_opt.py is done completely, scan_op.py
still has a bunch of PEP8 inconsistency.

After the fix I only run the tests in scan_module/tests.py and they passed
so I assume I had not introduced any bug in my PEP8 fixes.

Note also a test was added for this case.

Fixed typo in opt documentation

Also completed this doc and included a minor PEP8 fix

bad result are outputed. I have see that only in compyte and I was not able to make a test reproduice the error. The error appear with strieded inputs.

No initial inplace