Fix a case of advance indexing if the list of indices is not sorted.

[boazmohar]

• This now evaluates to true:

x = arange(3*3*4).reshape((3, 3, 4))
b = array(x, sc, axis=0)
assert allclose(b[[1, 0], [0, 1], [0, 2]].toarray(), x[[1, 0], [0, 1], [0, 2]])

• Mixed indexing already handled this case. I added more tests.
• Still can't handle sorting keys with dim > 1:
  This test passes:

b = array(x, sc, axis=(0,1))
assert allclose(b[[0, 1], [0, 1], [0, 2]].toarray(), x[[0, 1], [0, 1], [0, 2]])

This will not:

b = array(x, sc, axis=(0,1))
assert allclose(b[[1, 0], [0, 1], [0, 2]].toarray(), x[[1, 0], [0, 1], [0, 2]])

[boazmohar]

@jwittenbach Can you go over this?
Thanks!

[jwittenbach]

@boazmohar sorry for taking so long to get to this. It's looking good to me.

Do we have an open issue that it fixes (no worries if we don't)? I'm assuming that, without this fix, the test you wrote would fail when the indices are not ordered, correct? If so, is this just because it did not set the _sorted property correctly?

[boazmohar]

@jwittenbach no there is no open issue, and it is hard for me to remeber now but you also told me to get the order of the keys and not assume they are sorted: