Transposition is a concept of linear algebra rather than
multidimensional arrays of data.

`RowVector` is now defined as the `transpose` of any `AbstractVector`. If
`v` is an `AbstractVector`, then it obeys the identity that `(v.').'
=== v` and the matrix multiplication rules follow that `(A * v).' == (v.' *
A.')`. `RowVector` is a "view" and maintains the recursive nature of
`transpose`. It is a subtype of `AbstractMatrix` and maintains the
current shape and broadcast behavior for `v.'.

Consequences include:

 * v'' is a vector, not a matrix
 * v'*v is a scalar, not a vector
 * v*v' is the outer produce (returns a matrix)
 * v*A (for A::AbstractMatrix) is removed, since its puprose was to
   provide a way of doing the outer product above, and is no longer
   necessary.

Closes JuliaLang#4774