Deflate

* Fixed up stdcall support on 32-bit Windows: the functions are now exported
using both suffixed and non-suffixed names, and fixed `libdeflate.h` to be
MSVC-compatible again.

* The 32-bit Windows build of libdeflate now uses the "stdcall" calling
convention instead of "cdecl". If you're calling `libdeflate.dll` directly
from C or C++, you'll need to recompile your code. If you're calling it from
another language, or calling it indirectly using `LoadLibrary()`, you'll need
to update your code to use the stdcall calling convention.

* The Makefile now supports building libdeflate as a shared
library (`.dylib`) on macOS.

* Fixed a bug where support for certain optimizations and optional features
(file access hints and more precise timestamps) was incorrectly omitted when
libdeflate was compiled with `-Werror`.

* Added `make check` target to the Makefile.

* Added CI configuration files.

* `make install` now supports customizing the directories into which binaries,
headers, and libraries are installed.

* `make install` now installs into `/usr/local` by default. To change it, use
e.g. `make install PREFIX=/usr`.

* `make install` now works on more platforms.

* The Makefile now supports overriding the optimization flags.

* The compression functions now correctly handle an output data buffer >= 4 GiB
in size, and `gzip` and `gunzip` now correctly handle multi-gigabyte files (if
enough memory is available).

* Slight improvements to decompression speed.

* Added an AVX-512BW implementation of Adler-32.

* The Makefile now supports a user-specified installation `PREFIX`.

* Fixed build error with some Visual Studio versions.

* Fixed crash in CRC-32 code when the prebuilt libdeflate for 32-bit Windows was
called by a program built with Visual Studio.

* Improved the worst-case decompression speed of malicious data.

* Fixed build error when compiling for an ARM processor without hardware
floating point support.

* Improved performance on the PowerPC64 architecture.

* Added soname to `libdeflate.so`, to make packaging easier.

* Added `make install` target to the Makefile.

* The Makefile now supports user-specified `CPPFLAGS`.

* The Windows binary releases now include the import library for
`libdeflate.dll`. `libdeflate.lib` is now the import library, and
`libdeflatestatic.lib` is the static library.

* Added support for multi-member gzip files.

* Moved architecture-specific code into subdirectories. If you aren't using the
provided Makefile to build libdeflate, you now need to compile `lib/*.c` and
`lib/*/*.c` instead of just `lib/*.c`.

* Added an ARM PMULL implementation of CRC-32, which speeds up gzip compression
and decompression on 32-bit and 64-bit ARM processors that have the
Cryptography Extensions.

* Improved detection of CPU features, resulting in accelerated functions being
used in more cases. This includes:

* Detect CPU features on 32-bit x86, not just 64-bit as was done previously.

* Detect CPU features on ARM, both 32 and 64-bit. (Limited to Linux only
currently.)