Route similar_operator through TensorAlgebra.similar_map

Centralizes the linear-map allocation primitive in TensorAlgebra so backends only need to overload `similar_map`, while `similar_operator` keeps the named-axis wrapping.

Also drops the redundant `[sources.TensorAlgebra]` pin from `docs/`, `examples/`, and `test/Project.toml`; the workspace root `Project.toml` already pins the upstream feature branch for all subprojects.

Author: mtfishman

docs/Project.toml

@@ -9,10 +9,6 @@ Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 [sources.NamedDimsArrays]
 path = ".."
 
-[sources.TensorAlgebra]
-rev = "mf/gram-eigh-balanced"
-url = "https://github.com/ITensor/TensorAlgebra.jl"
-
 [compat]
 Documenter = "1"
 ITensorFormatter = "0.2.27"

examples/Project.toml

@@ -6,10 +6,6 @@ Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 [sources.NamedDimsArrays]
 path = ".."
 
-[sources.TensorAlgebra]
-rev = "mf/gram-eigh-balanced"
-url = "https://github.com/ITensor/TensorAlgebra.jl"
-
 [compat]
 NamedDimsArrays = "0.15"
 TensorAlgebra = "0.8, 0.9"

src/nameddimsoperator.jl

@@ -271,7 +271,7 @@ function similar_operator(
     unnamed_codomain_axes = conj.(unnamed_domain_axes)
     codomain_axes = named.(unnamed_codomain_axes, codomain_names)
     domain_axes = named.(unnamed_domain_axes, domain_names)
-    raw = similar(prototype, T, (codomain_axes..., domain_axes...))
+    raw = TA.similar_map(prototype, T, codomain_axes, domain_axes)
     return operator(raw, codomain_names, domain_names)
 end
 function similar_operator(

test/Project.toml

@@ -20,10 +20,6 @@ VectorInterface = "409d34a3-91d5-4945-b6ec-7529ddf182d8"
 [sources.NamedDimsArrays]
 path = ".."
 
-[sources.TensorAlgebra]
-rev = "mf/gram-eigh-balanced"
-url = "https://github.com/ITensor/TensorAlgebra.jl"
-
 [compat]
 AbstractTrees = "0.4.5"
 Adapt = "4"