Precision matrix estimates are non zero at zero locations

The function `fit_precision_cholesky` returns non-zero precision estimates in locations where the sparsity graph G say precision should be zero. From what I understand this is a bug. I have not figured out where the bug is yet. Small reproducible example:

```python3
import networkx as nx
import numpy as np

rng = np.random.default_rng(8)
n = 4 # Size
density = 0.4 # Density

# Create G indicating sparsity pattern
G = rng.uniform(size=(n, n)) < (density / 2)
G = G.T + G
np.fill_diagonal(G, np.ones(n))
G_mat = (G > 0).astype(int)
print("Sparsity pattern")
print(G_mat)
assert np.all(np.linalg.svd(G_mat).S > 0)
G = nx.from_scipy_sparse_array(sp.csc_array(G_mat))

# Create data U
U = rng.normal(size=(999, n))

# Estimate precision
Prec_est, *_ = fit_precision_cholesky(U=U, Graph_u=G, use_tqdm=False)
Prec_est = Prec_est.todense()
print("Estimated precision matrix")
print(Prec_est.round(4)) # Estimated precision matrix

# Look for non-zero precision where the sparsity pattern say the should be zero
entries_at_zero = Prec_est[G_mat == 0]
print("Entries at zero:")
entr= entries_at_zero[~np.isclose(entries_at_zero, 0.0)]
print(entr)
print(f"Number of entries at zero: {len(entr)}")
```

```text
Sparsity pattern
[[1 0 1 1]
[0 1 0 1]
[1 0 1 0]
[1 1 0 1]]
Estimated precision matrix
[[ 0.9576 0.0352 -0.0156 -0.0098]
[ 0.0352 0.9629 0. 0.0131]
[-0.0156 0. 0.9723 0. ]
[-0.0098 0.0131 0. 0.9614]]
Entries at zero:
[0.03518255 0.03518255]
Number of entries at zero: 2
```

Start by locating `fit_precision_cholesky` and run the reproducible example from the issue to inspect how the graph sparsity pattern is applied. Trace the precision estimate at entries where `G_mat == 0`; done means those entries remain zero while the existing estimation behavior is preserved.

Written by the indexing model from the issue text.