NaN and inf with nalgebra::linalg::SymmetricEigen

Hi,

I have a weird behaviour using SymmetricEigen. In my algorithm I want to diagonalize a covariance matrix, but it produces NaN and inf eigenvalues.
What is weird is when I try to reproduce the same behaviour printing my matrix using the same precision for the float values, SymmetricEigen seems to work well and give the right decomposition.

I double-checked my result using numpy and matlab

Here is the piece of code which tries to reproduct the weird behaviour without any success :

```
extern crate nalgebra;
use nalgebra as na;

fn main() {
let m = na::DMatrix::from_row_slice(10,10,
&[ 0.29094839096069335937500000000000, 0.0, 0.00125074386596679687500000000000, 0.0, 0.0, 0.0, 0.0, 0.23951797187328338623046875000000, 0.0, -0.53172069787979125976562500000000,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.00125074386596679687500000000000, 0.0, 0.00000604706247031572274863719940, 0.0, 0.0, 0.0, 0.0, 0.00115719600580632686614990234375, 0.0, -0.00241400394588708877563476562500,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
0.23951797187328338623046875000000, 0.0, 0.00115719600580632686614990234375, 0.0, 0.0, 0.0, 0.0, 0.22160133719444274902343750000000, 0.0, -0.46227973699569702148437500000000,
0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
-0.53172069787979125976562500000000, 0.0, -0.00241400394588708877563476562500, 0.0, 0.0, 0.0, 0.0, -0.46227973699569702148437500000000, 0.0, 0.99642109870910644531250000000000]
);
println!("The matrix {}", m);
let eigen_decomposition = na::linalg::SymmetricEigen::new(m);
println!("Eigenvalues : {}", eigen_decomposition.eigenvalues);

}
```

Thanks for your future answers or hints on what happen there.