Add format_descriptor<> & npy_format_descriptor<> PyObject * specializations. (#4674)

* Add `npy_format_descriptor<PyObject *>` to enable `py::array_t<PyObject *>` to/from-python conversions.

* resolve clang-tidy warning

* Use existing constructor instead of adding a static method. Thanks @Skylion007 for pointing out.

* Add `format_descriptor<PyObject *>`

Trivial addition, but still in search for a meaningful test.

* Add test_format_descriptor_format

* Ensure the Eigen `type_caster`s do not segfault when loading arrays with dtype=object

* Use `static_assert()` `!std::is_pointer<>` to replace runtime guards.

* Add comments to explain how to check for ref-count bugs. (NO code changes.)

* Make the "Pointer types ... are not supported" message Eigen-specific, as suggested by @Lalaland. Move to new pybind11/eigen/common.h header.

* Change "format_descriptor_format" implementation as suggested by @Lalaland. Additional tests meant to ensure consistency between py::format_descriptor<>, np.array, np.format_parser turn out to be useful only to highlight long-standing inconsistencies.

* resolve clang-tidy warning

* Account for np.float128, np.complex256 not being available on Windows, in a future-proof way.

* Fully address i|q|l ambiguity (hopefully).

* Remove the new `np.format_parser()`-based test, it's much more distracting than useful.

* Use bi.itemsize to disambiguate "l" or "L"

* Use `py::detail::compare_buffer_info<T>::compare()` to validate the `format_descriptor<T>::format()` strings.

* Add `buffer_info::compare<T>` to make `detail::compare_buffer_info<T>::compare` more visible & accessible.

* silence clang-tidy warning

* pytest-compatible access to np.float128, np.complex256

* Revert "pytest-compatible access to np.float128, np.complex256"

This reverts commit e9a289c50f.

* Use `sizeof(long double) == sizeof(double)` instead of `std::is_same<>`

* Report skipped `long double` tests.

* Change the name of the new `buffer_info` member function to `item_type_is_equivalent_to`. Add comment defining "equivalent" by example.

* Change `item_type_is_equivalent_to<>()` from `static` function to member function, as suggested by @Lalaland

tests/test_buffers.py

@@ -10,6 +10,63 @@ from pybind11_tests import buffers as m
 
 np = pytest.importorskip("numpy")
 
+if m.long_double_and_double_have_same_size:
+    # Determined by the compiler used to build the pybind11 tests
+    # (e.g. MSVC gets here, but MinGW might not).
+    np_float128 = None
+    np_complex256 = None
+else:
+    # Determined by the compiler used to build numpy (e.g. MinGW).
+    np_float128 = getattr(np, *["float128"] * 2)
+    np_complex256 = getattr(np, *["complex256"] * 2)
+
+CPP_NAME_FORMAT_NP_DTYPE_TABLE = [
+    ("PyObject *", "O", object),
+    ("bool", "?", np.bool_),
+    ("std::int8_t", "b", np.int8),
+    ("std::uint8_t", "B", np.uint8),
+    ("std::int16_t", "h", np.int16),
+    ("std::uint16_t", "H", np.uint16),
+    ("std::int32_t", "i", np.int32),
+    ("std::uint32_t", "I", np.uint32),
+    ("std::int64_t", "q", np.int64),
+    ("std::uint64_t", "Q", np.uint64),
+    ("float", "f", np.float32),
+    ("double", "d", np.float64),
+    ("long double", "g", np_float128),
+    ("std::complex<float>", "Zf", np.complex64),
+    ("std::complex<double>", "Zd", np.complex128),
+    ("std::complex<long double>", "Zg", np_complex256),
+]
+CPP_NAME_FORMAT_TABLE = [
+    (cpp_name, format)
+    for cpp_name, format, np_dtype in CPP_NAME_FORMAT_NP_DTYPE_TABLE
+    if np_dtype is not None
+]
+CPP_NAME_NP_DTYPE_TABLE = [
+    (cpp_name, np_dtype) for cpp_name, _, np_dtype in CPP_NAME_FORMAT_NP_DTYPE_TABLE
+]
+
+
+@pytest.mark.parametrize(("cpp_name", "np_dtype"), CPP_NAME_NP_DTYPE_TABLE)
+def test_format_descriptor_format_buffer_info_equiv(cpp_name, np_dtype):
+    if np_dtype is None:
+        pytest.skip(
+            f"cpp_name=`{cpp_name}`: `long double` and `double` have same size."
+        )
+    if isinstance(np_dtype, str):
+        pytest.skip(f"np.{np_dtype} does not exist.")
+    np_array = np.array([], dtype=np_dtype)
+    for other_cpp_name, expected_format in CPP_NAME_FORMAT_TABLE:
+        format, np_array_is_matching = m.format_descriptor_format_buffer_info_equiv(
+            other_cpp_name, np_array
+        )
+        assert format == expected_format
+        if other_cpp_name == cpp_name:
+            assert np_array_is_matching
+        else:
+            assert not np_array_is_matching
+
 
 def test_from_python():
     with pytest.raises(RuntimeError) as excinfo: