Added docs for the StridedArrays and fixed a doctest fail

[kshyatt]

I wasn't removing the file in that base/file doctest I wrote and it was bad. Now it is fixed. Also more docs for the Strided* types. Sorry for the bad branch name.

[kshyatt]

Review pls?

[mbauman]

How about s/may/might? "May not" sounds like it must not be contiguous.

I think it'd be good to introduce what strides are first. This is a little tricky. Here's one shot:

A union of common builtin N-dimensional array types with elements of type T that follow the strided array interface

If A isa StridedArray, then its entries are stored in memory with fixed offsets (or "strides") such that incrementing an index along dimension d jumps ahead in memory by strides(A, d) slots. Strided arrays are particularly important because they can sometimes be passed directly as pointers to foreign language libraries like BLAS.

Also note that strides of 1 isn't sufficient for contiguousness. The condition is: strides(A, 1) == 1 && strides(A)[2:end] == size(A)[1:end-1], but I don't think we need to talk about contiguousness here.

[kshyatt]

Finally got around to updating this