...fix RapidJSON dependency, remove features from CMake

I've now implemented in Meson all the remaining features that were in
CMake, like Cppcheck and Doxygen documentation. I've also removed almost
all the non-essential features from CMake, like building examples,
documentation, code coverage, the install target, Cppcheck, and tarball
generation. This should help with maintainability, while preserving
testability, following @kiplingw and @dennisjenkins75 main concerns.

Cppcheck is not ran by default, but it is a build target; you can run it
with `meson compile -C builddir cppcheck`.

I've also made some little fixes and tweaks:

Fixed RapidJSON
dependency, as it was previously declared only as a dependency of the
run_rest_description example, while used in
include/pistache/thirdparty/serializer/rapidjson.h.

Updated CMakeLists, replacing the old Include(Dart) with the newer
Include(CTest), and also adding messaging suggesting the user to install
RapidJSON and GoogleTest on their system rather than building them from
source.

Fixed a "small" issue in meson.build, replacing add_global_arguments
with add_project_arguments, since the former can't be used in
subprojects and can cause build failures for users.

Submodules get included in the source tarball generated by meson dist,
making dpkg-source throw a lot of warnings.

Using FetchContent instead of submodules has a few advantages:
Solves packaging warnings
Faster git clone
External dependencies get downloaded only if needed

Now the package is built with Meson. I've updated the debhelper-compat version to 12 since it provides good support for Meson, I've removed the lintian overrides (useless since the migration to Docusaurus), and obviously updated debian/rules to use Meson for building the package and creating the tarball.

This is not perfect though, as I had to use an ugly hack for the tarball since Meson does not add .github to the archive (mesonbuild/meson#8541).
Also, it is not possible to run the tests with valgrind since debhelper 12 uses ninja test for testing (while it should use meson test, so that one could add --wrap=valgrind). This behavior is corrected in debhelper 13, but it is too recent and not available in Ubuntu 20.04. It would be also possible to completely override dh_auto_test, but I don't think that's worth it.

autopkgtest does not report any error in Ubuntu Focal, so I think that everything is ok :)

As always, I forgot about something.

Upgraded to Ubuntu Focal, added Meson builds, added newer compilers

The problem was probably caused by
the absence of -lgcov as a linker option.
With this I'm replacing
-fprofile-arcs -ftest-coverage and the linker options
with --coverage, that passes all the correct flags
to compiler and linker.

Pistache now supports the Meson build system!

It is required to use Meson >=0.57
because previous versions can't read files,
and it wouldn't be possible to parse version.txt
to set the correct parameters in the .pc files and the shared library.
Such a recent version allows the users to
switch between the system-provided library
and the one they have in the subprojects folder
without even having to specify the fallback kwarg in dependency().

Since Meson provides builtin options for enabling code coverage
and PIC in static libraries there's no need for the PISTACHE_PIC option
and the "Profile target" mentioned by dennisjenkins75 in #840,
as they can be controlled with b_staticpic and b_coverage, respectively.

The cpp_std option is also set to c++17
even if the current minimun standard is C++14 because of #859.

If you don't know what Meson is, I highly recommend you to check it out.

The function was identical to parse_RFC_1123(), and the worst thing
is that tests with CMake weren't failing,
while running them with Meson revealed the issue.

"will additional functions" changing to "with additional functions"

pistache.io.yaml: Fixed typo.

Fix #863 issues

Bumped Ubuntu version for CI.

Docusaurus migration

listener: add CloseOnExec option

allows to set SOCK_CLOEXEC option for creating sockets
can be used to avoid leaking file descriptors to child processes
with fork()

Make fPIC optional

Added feature request template.

.github/ISSUE_TEMPLATE/bug_report.md: Removed parentheses around `about` stanza value...

Fixed bug report template.

Added bug report template.

Signed-off-by: Uilian Ries <uilianries@gmail.com>

Signed-off-by: Uilian Ries <uilianries@gmail.com>

Modernize CMakeLists.txt

WIth this I'm trying to fix some of the issues pointed out in #827.

Changes:

Use add_compile_options() instead of modifying CMAKE_CXX_FLAGS.

Use -Og instead of -O0
(from https://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html :
[-Og] is a better choice than -O0 for producing debuggable code because
some compiler passes that collect debug information are disabled at -O0)

Fix LTO check (previously it would fail because it checked only
CMAKE_CXX_FLAGS. This didn't work when properly enabling LTO
through CMAKE_INTERPROCEDURAL_OPTIMIZATION).
I'm not sure if I modified it correctly since I don't really understand
what that check is trying to do in the first place.

Check for Clang or GCC when applying compiler specific flags.

This changes require CMake 3.9 or higher, but I don't think this will
cause any issues.