Profiling via eBPF

[ashvardanian]

At this point we are sampling CPU, RAM and disk usage from Linux /proc/ files, just like perf.
There must be a better way with eBPF: intercepting instead of sampling.
Everyone is welcome to contribute.