Using fork1() with RFSTOPPED while being debugged with ptrace(PT_FOLLOW_FORK)
will cause the child to be forked off, never scheduled to attach debugger, and
the parent waiting for the child to attach at the end of do_fork() forever.
This should fix the hang with Linux fork(2) wrappers, which use RFSTOPPED, when
using ptrace(PT_FOLLOW_FORK). It also is much safer for other code that
may be using RFSTOPPED, such as vendor ABI modules, by removing the
possibility of the do_fork() code waiting forever for an unscheduled child.
I like the idea of the fork return hooks and still want to see that
implemented, but prefer this approach for this problem as otherwise the bug
is still present for other code to run into.