x86/vdso: Fake 32bit VDSO build on 64bit compile for vgetcpu

The 64bit register constrains in __arch_hweight64() cannot be
fulfilled in a 32-bit build. The function is only declared but not used
within vclock_gettime.c and gcc does not care. LLVM complains and
aborts. Reportedly because it validates extended asm even if latter
would get compiled out, see

  https://lore.kernel.org/r/Y%2BJ%2BUQ1vAKr6RHuH@dev-arch.thelio-3990X

i.e., a long standing design difference between gcc and LLVM.

Move the "fake a 32 bit kernel configuration" bits from vclock_gettime.c
into a common header file. Use this from vclock_gettime.c and vgetcpu.c.

  [ bp: Add background info from Nathan. ]

Fixes: 92d33063

 ("x86/vdso: Provide getcpu for x86-32.")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/Y+IsCWQdXEr8d9Vy@linutronix.de