py3: fixes in matrices

[jhpalmieri]

This fixes a few doctests for sage.matrices:

• there is some randomness in the ordering in one small part of one verbose test in compute_J_ideal.py, so replace it with ...
• up to now, mat[range(3), range(4)] did not work, so I've tried to modify __getitem__ and __setitem__ to allow range types as inputs
• some doctests use random matrices to test other features. The randomness differs on Python 2 vs. Python 3, so get rid of the randomness. The tests still test the relevant methods (which are determinants, for example, not the randomness).
• there is minor numerical noise in one doctest.

[jhpalmieri]

Regarding the random matrix issue: I think what's going on is that the different versions of Python are using different orderings on monomials in QQ.<x,y>, for instance, so after coming up with a list of random coefficients, the resulting polynomials are different. (With Python 2, you would see a*y^2 + b*x*y while in Python 3 you would see a*x*y + b*y^2, same coefficients a and b in the two cases.)

Regarding the changes to allow range(...) as valid input to __getitem__ and __setitem__ for matrices: are these changes okay? Is there a better way to do it?

---

New commits:

​888612c

trac 27585: work on Python 3 compatibility for sage.matrix

[jhpalmieri]

After this ticket, there is one remaining doctest failure in sage.matrix:

TypeError: type sage.rings.real_double.RealDoubleElement doesn't define __round__ method

in matrix_double_dense.pyx.

[vklein]

patchbot errors are unrelated with this ticket.
Looks good to me.