Symbolize from proc maps

[r1viollet]

• When pid option is specified, check all binaries from proc maps.
• Lazy symbolization
  Retrieve symbols when needed. Add a cache to check if symbols were loaded.

Authored by @nsavoire

[nsavoire]

Symbolization from /proc/%pid%/maps mostly works, but is not yet well integrated with the rest of the code.

Issues/ToDo:

• Remove TracedBinary object, it looks like it is not needed anymore
• User stack traces are printed as kernel stack traces (because TracedBinary is nil in StackTrace object), linked to the previous remark.
• Uprobe requires either a binary path prefix in the pattern or a pid. if an excutable is given, we could imagine finding pids running with this executable.
• If several pids are given, different funcIds might be created for the same binary / offset (one for each pid)
• There are 2 different mechanisms for PID filtering:
  • target a specific PID with perf API (only possible for uprobes/perfevents): minimal impact on non-traced PIDs but newly created processes can't be traced easily
  • use an EBPF map inside EBPF program to determine is current PID is traced (for system wide traces such as kprobes/tracepoints and also checked for uprobes/perfevents): runtime impact on non-traced PIDs but newly created processes are automatically added by tracepoint_sched_sched_process_exec and traced
• Current trigger for capturing user stack traces in kprobe/tracepoints/perf_events is filter_user_binary, not sure about this

[r1viollet]

@Gui774ume if you want to check out this PR. There are still some limitations, but I feel the mechanisms added are very interesting.