scx_rusty: Don't use PID to identify tasks between userspace and BPF

[htejun]

scx_rusty implements most of load balancing logic in userspace and uses PID to identify tasks between userspace and BPF. However, a task's PID can change when a non-leader thread exec(2)s as it swaps its PID with the group leader before killing all other threads and exec(2)ing. This PID change can lead to various malfunctions in scx_rusty including occasional crashes. #437 is likely one of the symptoms too.

It'd be useful to create a repro case which can demonstrate the problem reliably and fix it by changing the key to something actually persistent and unique - e.g. the task pointer.