ndarray.shape,strides act more like their python counterparts see #157

[mborgerding]

addresses #157
Unit tests added for to verify correct behavior for shape(k),strides(k) with negative indexing. Verify IndexError raised when k out of bounds.

[stefanseefeld]

The change itself looks good, but it shouldn't be lumped into the PR about negative strides and indices. (For avoidance of doubt: I'm talking about the second commit only. The first is fine, and only awaiting a completed test run on travis.)

[mborgerding]

@stefanseefeld , I agree the changes shouldn't be lumped together. That was not my intent, but github did it on its own (I may've missed something) . If you feel they should be separated, I can try again, maybe rebasing both changes from the master branch would force github to make them separate PRs.

[stefanseefeld]

Yeah, simply move the second one into a separate branch from which you can submit another PR. Sorry it's all a bit formal, but with practice it becomes easier ;-)

[mborgerding]

Split this into two PRs. From now on I'll know not to believe that github's "Create New Pull Request" will actually create a New pull request.

[stefanseefeld]

FWIW, you could simply have added another commit reverting the previous one (which I could have squash-and-merged as one - so only the first one would have been visible), or you could have force-pushed an update removing the second commit.
But of course, starting over with a new PR works, too. :-)