Bug: normalize fails with TypeError (boolean check in __normalize!)

[bdrhill]

Problem

LinearAlgebra.normalize fails with TypeError due to a boolean check in __normalize!.

MWE

using SparseConnectivityTracer
using LinearAlgebra

detector = TracerSparsityDetector()

# This fails
jacobian_sparsity(x -> normalize(x), [3.0, 4.0], detector)
# TypeError: non-boolean (GradientTracer...) used in boolean context

# Manual implementation works
jacobian_sparsity(x -> x / norm(x), [3.0, 4.0], detector)
# OK: nnz=4

Root cause

LinearAlgebra.__normalize! at generic.jl:2010 uses a boolean check (likely iszero on the norm).

Workaround

Use manual implementation x / norm(x) or TracerLocalSparsityDetector.