[py framework] Setting context values now clones instead of owns

[jwnimmer-tri]

For the following classes ...

• ContinuousState
• DiscreteValues
• Parameters

... when setting a new value from Python (either in the constructor, or with a set function) the value is now copied instead of aliased. Having the context keep mutable aliases to objects that the user provided is not only too dangerous in the first place, but also incompatible with forthcoming upstream pybind11 semantics.

Towards #21968.

---

This change is 

[jwnimmer-tri]

+@sherm1 for feature review, please.

[sherm1]

Feature with one question

Reviewed 2 of 2 files at r1, all commit messages.
Reviewable status: 1 unresolved discussion, needs at least two assigned reviewers

---

bindings/pydrake/systems/framework_py_semantics.cc line 1106 at r1 (raw file):

        return std::make_unique<ContinuousState<T>>(
            state.Clone(), num_q, num_v, num_z);
      }),

BTW do the BasicVector-matching overloads have unit tests somewhere?

[jwnimmer-tri]

+@rpoyner-tri for platform review, please.

Reviewable status: LGTM missing from assignee rpoyner-tri(platform)

---

bindings/pydrake/systems/framework_py_semantics.cc line 1106 at r1 (raw file):

Previously, sherm1 (Michael Sherman) wrote…

BTW do the BasicVector-matching overloads have unit tests somewhere?

They both have code coverage (tests do run them -- confirmed by inserting DRAKE_UNREACHABLE).

However, if I remove them leaving only the VectorBase, then all tests still pass. So, I added more tests for the BasicVector subclassing.

[sherm1]

Reviewed 2 of 2 files at r2, all commit messages.
Reviewable status: LGTM missing from assignee rpoyner-tri(platform)

[rpoyner-tri]

Reviewed 2 of 2 files at r1, 2 of 2 files at r2, all commit messages.
Reviewable status: complete! all discussions resolved, LGTM from assignees rpoyner-tri(platform),sherm1(platform)