The -Wmissing-prototypes Clang warning (or -Wmissing-declarations on
GCC) is very useful to avoid accidents where a function definition in a
source file doesn't match the corresponding declaration in a header
file, as it would warn already during compilation and not much later
during link time.

Unfortunately this means that exported functions defined only in the
source file (usually the ones annotated with `extern "C"`) will cause
this warning to be emitted too (on Clang, GCC has a slightly different
behavior with -Wmissing-declarations and doesn't warn here). This fixes
the warning by providing a declaration right before the definition.

Clang has a bug [1] in x86 Windows that is exposed by the use of lambdas with "unforwardable" prototypes. The error is "error: cannot compile this forwarded non-trivially copyable parameter yet", and the message was introduced in [2] (used to be an assertion).

[1] https://llvm.org/bugs/show_bug.cgi?id=28299
[2] https://github.com/microsoft/checkedc-clang/commit/feb1567e07573100ea14f9aea02f81463e791496

* In internals.h, only look at _DEBUG when compiling with MSVC.

(_DEBUG is a MSVC-specific macro.)

PyArray_DescrFromType returns a new reference, not borrowed one

Test for dtype checks now succeed without warnings

numpy: Add test for explicit dtype checks. At present, int64 + uint64 do not exactly match dtype(...).num

* Added ability to convert from Python datetime.date and datetime.time to C++ system_clock::time_point

* Protect std::min/max functions from windows.h min/max
Removed check for windows min/max

This is only necessary if `get_internals` is called for the first time in a given module when the running thread is in a GIL-released state.

Fixes #1364

* Test dtype field order in numpy dtype tests

When running tests with NumPy 1.14 or later this test exposes the
"invalid buffer descriptor" error reported in #1274.

* Create dtype_ptr with ordered fields

* Fix casting of time points with non-system-clock duration on Windows

Add explicit `time_point_cast` to time point with duration of system
clock. Fixes Visual Studio compile error.

* Add test case for custom time points casting

In def_readonly and def_readwrite, there is an assertion that the member comes
from the class or a base class:

    static_assert(std::is_base_of<C, type>::value, "...");

However, if C and type are the same type, is_base_of will still only be true
if they are the same _non-union_ type.  This means we can't define accessors
for the members of a union type because of this assertion.

Update the assertion to test

    std::is_same<C, type>::value || std::is_base_of<C, type>::value

which will allow union types, or members of base classes.

Also add a basic unit test for accessing unions.

* Fix async Python functors invoking from multiple C++ threads (#1587)

Ensure GIL is held during functor destruction.

* Add async Python callbacks test that runs in separate Python thread

In some cases the user of pythonbuf needs to allocate the internal
buffer to a specific size e.g. for performance or to enable synchronous
writes to the buffer.
By changing `pythonbuf::d_buffer` to be dynamically allocated we can now
enable these use-cases while still providing the default behavior of
allocating a 1024 byte internal buffer (through a default parameter).

fix some uninitialized members in `value_and_holder` for
some of the constructurs.

Found with coverity in a downstream project.

I found that the numpy array tests already contained an empty-shaped
array test, but none with data in it.

Following PEP 3118, scalars have an empty shape and ndim 0. This
works already and is now also documented/covered by a test.