The binary location had not been updated following the move of the dseries
submodule (commit e962cb4d).

[skip ci]

Incidentally, this means that the JSONlab submodule is no longer needed, so it
is removed in the present commit.

Closes: #1907

Ref. #1907

Incidentally, drop workarounds for pkg-config files of HDF5 and MatIO, since
the problems have been fixed upstream.

By the way, simplify the dependency build system by removing unused SLICOT
builds.

Mostly working. The testsuite remains to be done.

In particular, windows/deps/mkoctfile64 is modified to print its version number
on stdout instead of stderr. This is what the standard mkoctfile does, and it
allows easier capture of the output from Meson.

Unpacked files may already be present due to GitLab CI caching.

Bug introduced in 39c2cba1.

Ref. #1878

– enable maximum optimization (-O3), which in particular triggers code
  auto-vectorization
– raise the minimal architectural requirement to match MSYS2 policy
  (roughly requires CPUs from ~2006/7+)

In particular, now ship GCC 13 for the use_dll option.

[skip ci]

Also move some useful documentation to doc/.

HDF5 itself now drags a lot of new dependencies (curl and its own deps,
openssl).

For the time being, the HDF5 package is held back, because of this issue which
affects version 1.14.0:
https://github.com/msys2/MINGW-packages/issues/15449#issuecomment-1470098865

In particular, add -fstack-protector to CXXFLAGS, to avoid undefined symbols at
link time.

This shortcut is filtered out by Windows ⩾ 8, so it is useless.

As a consequence, only create a start menu entry when documentation is
installed.

Incidentally, no longer run Dynare++ unit tests, since they are intricated in
the Dynare++ build system. If we finally decide not to drop the Dynare++
codebase for our MEX files, then we should probably find a way to build and run
them again.

Ref. #1825