Let cond of an empty square matrix return zero

[martinholters]

Transfer of JuliaLang/julia#38372. Summary of the discussion there:
We have this inconsistency:

julia> cond(ones(0,0), 1)
0.0

julia> cond(ones(0,0), 2)
ERROR: ArgumentError: reducing over an empty collection is not allowed; consider supplying `init` to the reducer

julia> cond(ones(0,0), Inf)
0.0

If it wasn't breaking, making all cases throw would be acceptable behavior. This PR goes the other way, making the p==2 case also return zero. Although met with some skepticism, if there is to be chosen a value, zero is probably it.

I think the main question is whether the inconsistency or the somewhat dubious return-value is the lesser evil. Mainly porting this over as a reminder to reach a decision. Either way we should probably close #778; either by resolving it with this PR or as won't-fix.

Fixes #778.

[stevengj]

Thinking about this a bit more, this seems sensible, at least for square empty matrices. For square matrices, we define the condition number as $\Vert A \Vert \cdot \Vert A^{-1} \Vert$. For an empty matrix, the norm should be zero (because it's the additive identity), and the inverse of a square empty matrix is itself (since $Ax = b$ has a unique solution: an empty $b$ is unique, and empty $x$ solves it).

Returning 0 is also what Matlab does, it seems.

[martinholters]

Agree, restricting to square A is reasonable.

[stevengj]

Would be good to have a @test_throws to ensure it throws for non-square empty matrices.