ktrace: Fix some problems with process subtree traversal

[markj]

Summary

- Avoid visiting processes that were reparented, since they represent
debuggees possibly belonging to a different subtree.
- Visit orphan processes, to ensure that reparented debuggees that do
belong to the subtree are treated consistently.

Reported by: kib

[kib]

sys/kern/kern_ktrace.c

1244

I understand why this assert should be true for 'p' equal to the function argument. But after we started iterating over siblings, I think this is no longer true, generally.

1268

For orphans, I believe the iteration should be also done in depth, same as for siblings. Children of an orphan are also descendants of the process.