[py trajectories] Adjust Clone and MakeDerivative for unique_ptr

[jwnimmer-tri]

Towards #5842 and #21968.

The situation with Trajectory is a bit different that some of our other similar changes to bindings for those issues. In most other cases, the only trouble comes from Clone, but with Trajectory we also have an instance factory method MakeDerivative() that also needs a solution. So in this case, we'll use the decorator pattern to solve the ownership problem (via internal::WrappedTrajectory) instead of trying to deal with Clone directly.

---

This change is 

[jwnimmer-tri]

+@rpoyner-tri for feature review, please? This is a companion to #22353. Or maybe is there someone else who we could ramp up on bindings details like this to grow the feature reviewer pool?

[rpoyner-tri]

Reviewed 7 of 7 files at r1, all commit messages.
Reviewable status: 3 unresolved discussions, needs at least two assigned reviewers

---

bindings/pydrake/trajectories_py.cc line 205 at r1 (raw file):

            captured_py_traj = py::none();
          });
      // Wrap it in a unique_ptr to meet out required return signature.

typo

Suggestion:

our

---

bindings/pydrake/trajectories_py.cc line 205 at r1 (raw file):

            captured_py_traj = py::none();
          });
      // Wrap it in a unique_ptr to meet out required return signature.

nit This comment doesn't quite do it for me. My first reading missed the crucial role of WrappedTrajectory in this particular caper.

Maybe instead: "Stuff the shared pointer inside a WrappedTrajectory, and return that via unique_ptr to meet our required return signature."

---

bindings/pydrake/trajectories_py.cc line 223 at r1 (raw file):

      bool used_legacy_clone = true;
      auto make_python_deepcopy = [&]() -> py::object {
        PYBIND11_OVERLOAD_INT(py::object, Trajectory<T>, "Clone");

nit similar to elsewhere, it's probably useful to point out that this macro lets control fall through if there is no suitable binding.

[rpoyner-tri]

Reviewed 1 of 1 files at r2, all commit messages.
Reviewable status: 2 unresolved discussions, needs at least two assigned reviewers

[rpoyner-tri]

Reviewed 1 of 1 files at r3, all commit messages.
Reviewable status: needs at least two assigned reviewers

[jwnimmer-tri]

+@sherm1 for off-schedule platform review, please. Looking for some fresh eyes on this as the on-call people this week are already primed with a lot of background knowledge.

Reviewed 6 of 7 files at r1, 1 of 1 files at r3, all commit messages.
Reviewable status: LGTM missing from assignee sherm1(platform)

---

bindings/pydrake/trajectories_py.cc line 205 at r1 (raw file):

Previously, rpoyner-tri (Rick Poyner (rico)) wrote…

nit This comment doesn't quite do it for me. My first reading missed the crucial role of WrappedTrajectory in this particular caper.

Maybe instead: "Stuff the shared pointer inside a WrappedTrajectory, and return that via unique_ptr to meet our required return signature."

"caper" -- I love it.