Swapping Dimensions

[Tokazama]

• [ ] Finalize design:
  • [x] permutedims
  • [x] inv
  • [ ] cov/cor
    • [x] dims= 1/2
    • [ ] Any other dimension gives a diagonal with NaNs. The solution right now uses the same method for diag (described below) but then uses that axis for the first and second axes of the returned matrix.
  • [x] Matrix multiplication
  • [ ] diag: the current solution is equivalent to promoting both axes (promote(axes(m)...)) and returning the shortest one (kind of like the rank axis).
• [x] Document axis interface
• [ ] Document customization with axes
• [ ] Examples with AbstractAxisIndices

Most of these are very straightforward. I feel like the current solution for the deriving the axis that corresponds to the diagonal is pretty good but I'm not as set on it as the others.

[Tokazama]

v0.2.0 consolidated inverse_axes into permute_axes. Doesn't change behavior just simplifies code.