fix bug in issymmetric/ishermitian when zeros are stored

[KristofferC]

Fixes #13645

I introduced a small bug in #11371 that made issymmetric give the wrong answer in some cases when the matrix had stored zeros.

In short, the problem was that when I skipped storedzeros I have take into account that the tracker that keeps track of the next element needs to be incremented as well.
I took the opportunity to add some comments to the code as well.

With:

m = n = 5
colptr = [1, 5, 9, 13, 13, 17]
rowval = [1, 2, 3, 5, 1, 2, 3, 5, 1, 2, 3, 5, 1, 2, 3, 5]
nzval = [0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0, 0.0, 1.0, 1.0, 1.0]
A = SparseMatrixCSC(m, n, colptr, rowval, nzval)

We have:

julia> A-A'
5x5 sparse matrix with 0 Float64 entries:

On master:

julia> issymmetric(A) 
false

On PR:

julia> issymmetric(A) 
true

[tkelman]

use a let here?

[KristofferC]

Thank you @tkelman. Updated.