0.13.2 --- a thread created from C++ is the thread that is joined

[Sunrisepeak]

Every std::thread on Windows ended in an access violation when it was joined.

musl declares pthread_t twice — a pointer for C, unsigned long for C++ — and
the two agree wherever a long holds a pointer. Windows is LLP64, so C++ code
above this library kept the lower half of the thread's address. libc++ stores
pthread_t in std::thread, and pthread_join read through the truncated value:

wine: Unhandled page fault on read access to 00000000FE9B8B60 at address 000000014003B2BA
  -> pthread_timedjoin_np, musl/src/thread/pthread_join.c:16
     rax=00000000fe9b8b28   (the thread pointer, upper half gone)

On windows-2022 the same programs exit with 0xC0000005. Found through
lsp-mcpp's unit tests, whose thread and process tests crashed only on Windows.

• musl/include/alltypes.h.in and the five generated alltypes.h: C++
  pthread_t is unsigned _Addr. _Addr is what musl already uses for size_t
  and uintptr_t; it is long on x86_64, aarch64 and riscv64, so nothing
  changes there, and long long on x86_64-windows.
• musl/PATCHES.md records it as the fifth patched line, the same kind as the
  other four: a machine word carried through a long.
• examples/threads-cxx asserts the width with static_assert (it does not
  compile for x86_64-windows-gnu before this change) and creates and joins a
  thread; CI runs it on every row.

Measured locally: the example fails to compile for x86_64-windows-gnu before
the change; after it, the Windows build prints joined: 0, value 42, result is the argument under Wine. With openkal-llvm-runtime pointed at this tree, std::thread,
a condition variable across threads and an exception thrown and caught inside a
thread all work under Wine.

0.13.2. Correction after merge: openkal-llvm-runtime pins openkal-musl = "0.13.1"
exactly (a consumer asking for 0.13.2 beside it is refused as irreconcilable), so
consumers reach this fix through openkal-llvm-runtime 0.9.3
(mcpplibs/openkal-llvm-runtime#18).