cptbox multiprocess support

[quantum5]

cptbox, after #188, should logically implement multiprocess debugging support. This may one day permit the implementation of a BASH language, if we manage to limit what it can exec.

[quantum5]

Now, everywhere we use kill, i.e. shocker, protection faults, we have to kill by pgid, which for cptbox-based process trees is -pid of the first child. (Even though monitor does attempt to kill all children that it is aware of, it's still unwise to rely on the fact.)

[quantum5]

Debian/kFreeBSD jessie does not have PT_FOLLOW_FORK defined. However, passing 23 works.

[Riolku]

Isn't this complete?

[Xyene]

No, this is not implemented in a safe manner yet, or we'd have shell executors on DMOJ :)

[Riolku]

Will landlock allow this to be cleaned up?

[Xyene]

Yes, that's one of the biggest missing pieces.

[Riolku]

Would we need to only allow bash when paired with landlock? I guess so... is the other option to stop all processes when a file access happens?

[Xyene]

Yes, and ensure that they actually all get stopped, which sounds painful. We'd probably want to have some non-bash-builtins too, but this is super low-priority so hasn't been given much thought.