Python wrapper: prepare wrapped python objects for numpy usage

[akuesters]

see https://pybind11.readthedocs.io/en/stable/advanced/pycpp/numpy.html?highlight=numpy

Add buffer_info to all wrapped python objects returning lists (e.g. stdandard vector, list, array, deque)

[brenthuisman]

Probes are already covered.

[thorstenhater]

buffer_info is only needed where we define an object that is buffer-like (eg matrices), example
from https://pybind11.readthedocs.io/en/stable/advanced/pycpp/numpy.html#buffer-protocol

py::class_<Matrix>(m, "Matrix", py::buffer_protocol())
   .def_buffer([](Matrix &m) -> py::buffer_info {
        return py::buffer_info(m.raw,                                  /* Pointer to buffer */
                               sizeof(float),                          /* Size of one scalar */
                               py::format_descriptor<float>::format(), /* Python struct-style format descriptor */
                               2,                                      /* Number of dimensions */
                               {m.rows(), m.cols()},                   /* Buffer dimensions */
                               {sizeof(float)*m.cols(),                /* Strides (in bytes) for each index */
                                sizeof(float) });
    });

Everywhere else, we need return a pybind11::array or pybind11::array_t<T> for numpy
types and pybind11::buffer for more general buffer-enabled classes.

The 0-copy way of doing this is the following

// Get a pointer to the vector object (this is copy-free, since move-constructed)
auto v = new std::vector<int>(some_func());
// pybind11 smart-pointer any hybrid with explicit deleter
auto capsule = py::capsule(v, [](void *v) { delete reinterpret_cast<std::vector<int>*>(v); });
// Create an array from the raw data and capsule, where the capsule's lifetime is bound
// to that of the array and in turn will (C++) delete the vector object.
return py::array(v->size(), v->data(), capsule);

if we can live with a copy, this should be used to be less stressful on my sanity (no lifetime management madness)

auto v = some_func();
return py::array(v.size(), v.data());

See pybind/pybind11#1042

As @halfflat say, the probe API is covered already and with that the main usecase. After trawling the Python
interface I could not find more, so I am closing this with the request to re-open, if I missed something.
The notes above can serve as reference on how to achieve numpy compatibility where needed.