I am checking here before opening an issue.

Using the above jld file holding a symmetric matrix A, the following results are very weird:

```
julia> d = load("posdef_symm_sparse_matrix.jld");
julia> A = d["A"];
julia> typeof(A)
Symmetric{Float64,SparseMatrixCSC{Float64,Int64}}
julia> isposdef(A.data)
false
julia> isposdef(Array(A.data))
false
julia> isposdef(Array(A))
true
julia> Array(A) ≈ Array(A.data)
true
julia> Array(A) == Array(A.data)
false
julia> min(eigvals(Array(A.data))...)
0.00017220258699615022
julia> min(eigvals(Array(A))...)
0.0001722025869961005
```

Is there a more reliable way to check if a sparse matrix is positive definite or not in v0.6.2? The `eigmin`

function fails on sparse matrices in v0.6.2. And finding the whole eigenvalue decomposition is overkill.