Fix inconsistent pybind11/eigen/tensor.h behavior:

This existing comment in pybind11/eigen/tensor.h

```
                // move, take_ownership don't make any sense for a ref/map:
```

is at odds with the `delete src;` three lines up.

In real-world client code `take_ownership` will not exist (unless the client code is untested and unused). I.e. the `delete` is essentially only useful to avoid leaks in the pybind11 unit tests.

While upgrading to clang-tidy 18, the warning below appeared. Apparently it is produced during LTO, and it appears difficult to suppress. Regardless, the best way to resolve this is to remove the `delete` and to simply make the test objects `static` in the unit test code.

________

Debian clang version 18.1.8 (++20240718080534+3b5b5c1ec4a3-1~exp1~20240718200641.143)
Target: x86_64-pc-linux-gnu
________

```
lto-wrapper: warning: using serial compilation of 3 LTRANS jobs
lto-wrapper: note: see the ‘-flto’ option documentation for more information
In function ‘cast_impl’,
    inlined from ‘cast’ at /mounted_pybind11/include/pybind11/eigen/tensor.h:414:25,
    inlined from ‘operator()’ at /mounted_pybind11/include/pybind11/eigen/../pybind11.h:296:40,
    inlined from ‘_FUN’ at /mounted_pybind11/include/pybind11/eigen/../pybind11.h:267:21:
/mounted_pybind11/include/pybind11/eigen/tensor.h:475:17: warning: ‘operator delete’ called on unallocated object ‘<anonymous>’ [-Wfree-nonheap-object]
  475 |                 delete src;
      |                 ^
/mounted_pybind11/include/pybind11/eigen/../pybind11.h: In function ‘_FUN’:
/mounted_pybind11/include/pybind11/eigen/../pybind11.h:297:75: note: declared here
  297 |                     std::move(args_converter).template call<Return, Guard>(cap->f),
      |                                                                           ^
```

Author: rwgk

include/pybind11/eigen/tensor.h

@@ -469,9 +469,6 @@ struct type_caster<Eigen::TensorMap<Type, Options>,
                 parent_object = reinterpret_borrow<object>(parent);
                 break;
 
-            case return_value_policy::take_ownership:
-                delete src;
-                // fallthrough
             default:
                 // move, take_ownership don't make any sense for a ref/map:
                 pybind11_fail("Invalid return_value_policy for Eigen Map type, must be either "

tests/test_eigen_tensor.inl

@@ -147,33 +147,39 @@ void init_tensor_module(pybind11::module &m) {
 
     m.def(
         "take_fixed_tensor",
-
         []() {
             Eigen::aligned_allocator<
                 Eigen::TensorFixedSize<double, Eigen::Sizes<3, 5, 2>, Options>>
                 allocator;
-            return new (allocator.allocate(1))
+            static auto *obj = new (allocator.allocate(1))
                 Eigen::TensorFixedSize<double, Eigen::Sizes<3, 5, 2>, Options>(
                     get_fixed_tensor<Options>());
+            return obj; // take_ownership will fail.
         },
         py::return_value_policy::take_ownership);
 
     m.def(
         "take_tensor",
-        []() { return new Eigen::Tensor<double, 3, Options>(get_tensor<Options>()); },
+        []() {
+            static auto *obj = new Eigen::Tensor<double, 3, Options>(get_tensor<Options>());
+            return obj; // take_ownership will fail.
+        },
         py::return_value_policy::take_ownership);
 
     m.def(
         "take_const_tensor",
         []() -> const Eigen::Tensor<double, 3, Options> * {
-            return new Eigen::Tensor<double, 3, Options>(get_tensor<Options>());
+            static auto *obj = new Eigen::Tensor<double, 3, Options>(get_tensor<Options>());
+            return obj; // take_ownership will fail.
         },
         py::return_value_policy::take_ownership);
 
     m.def(
         "take_view_tensor",
         []() -> const Eigen::TensorMap<Eigen::Tensor<double, 3, Options>> * {
-            return new Eigen::TensorMap<Eigen::Tensor<double, 3, Options>>(get_tensor<Options>());
+            static auto *obj
+                = new Eigen::TensorMap<Eigen::Tensor<double, 3, Options>>(get_tensor<Options>());
+            return obj; // take_ownership will fail.
         },
         py::return_value_policy::take_ownership);