framework_py: Add test of using AbstractValue's in input ports, fix overload error for unique_ptr in pybind

[EricCousineau-TRI]

Closes #8074

Address part of AbstractValue point in #7660. (Just need to follow-up with unrestricted updates.)

When I was first prototyping accepting unique_ptr as arguments in pybind, I introduced a bug where any registered pybind type would match with any other pybind type. This is fixed via RobotLocomotion/pybind11#10
(I ultimately made this a possibility when supporting anti-slicing stuff for shared_ptr, but just hadn't incorporated it until now...)

This also fixes a minor issue with cpp_template - any instantiations may have had the wrong module name used because inspect was not called with the right frame count. That has now been fixed.

---

This change is 

[EricCousineau-TRI]

Fixed the random errors in the pybind PR (due to recycled memory, instance registration stuff).

+@jadecastro for feature review, please.
+@sammy-tri for platform review, please.

---

Review status: 0 of 7 files reviewed at latest revision, all discussions resolved.

---

Comments from Reviewable

[EricCousineau-TRI]

Pybind PR has landed - updated in pybind/repository.bzl.

[jadecastro]

Pass complete.

---

Reviewed 7 of 7 files at r1.
Review status: all files reviewed at latest revision, 3 unresolved discussions, some commit checks failed.

---

bindings/pydrake/systems/framework_py.cc, line 283 at r1 (raw file):

        [](const System<T>* self, const Context<T>& arg1, int arg2) {
          return self->EvalVectorInput(arg1, arg2);
        }, py_reference_internal)

BTW, comparing with the abstract binding, it no longer makes sense to me for this one to be a py_reference_internal, but instead a py_reference, py::keep_alive<0, 2>() like below since it's the lifespan of the context we want to control?

---

bindings/pydrake/systems/primitives_py.cc, line 36 at r1 (raw file):

  py::class_<PassThrough<T>, LeafSystem<T>>(m, "PassThrough")
    .def(py::init<int>())

BTW, could you also check/add a test for this (vector) version of the ctor?

---

bindings/pydrake/systems/test/general_test.py, line 172 at r1 (raw file):

    def test_abstract_pass_through(self):
        model_value = AbstractValue.Make("Hello world")
        input_value = model_value.Clone()

BTW, I don't think this needs to be cloned as a separate value.

---

Comments from Reviewable

[sammy-tri]

(pending existing issues/cleanup)

---

Reviewed 7 of 7 files at r1.
Review status: all files reviewed at latest revision, 3 unresolved discussions, some commit checks failed.

---

Comments from Reviewable

[EricCousineau-TRI]

Review status: 2 of 7 files reviewed at latest revision, 3 unresolved discussions.

---

bindings/pydrake/systems/framework_py.cc, line 283 at r1 (raw file):

Previously, jadecastro (Jonathan DeCastro) wrote…

BTW, comparing with the abstract binding, it no longer makes sense to me for this one to be a py_reference_internal, but instead a py_reference, py::keep_alive<0, 2>() like below since it's the lifespan of the context we want to control?

Done. Good catch - thanks!

---

bindings/pydrake/systems/primitives_py.cc, line 36 at r1 (raw file):

Previously, jadecastro (Jonathan DeCastro) wrote…

BTW, could you also check/add a test for this (vector) version of the ctor?

Done.

---

bindings/pydrake/systems/test/general_test.py, line 172 at r1 (raw file):

Previously, jadecastro (Jonathan DeCastro) wrote…

BTW, I don't think this needs to be cloned as a separate value.

Done.

---

Comments from Reviewable

[jadecastro]

---

Reviewed 5 of 5 files at r2, 1 of 1 files at r3.
Review status: all files reviewed at latest revision, all discussions resolved.

---

Comments from Reviewable

[EricCousineau-TRI]

Deferring merge until #8110 lands.