ptrace: syscall-exit && stepping fixes

Currently PTRACE_SINGLESTEP after syscall_exit stop is not consistent across different machines, and we have different problems depending on arch/.

As Roland suggests, we should move the arch-dependent logic into the tracehook_report_syscall_exit() and unify/fix the behaviour.

Oleg.

At least 4-5/5 should be CC'd to the x86 maintainers too. At least 1/5 should perhaps go to linux-arch to alert other arch maintainers that implement user_enable_single_step() to add user_single_step_siginfo() for their arch too.

Thanks,
Roland