refactoring some code (stats, writers)

[xaki23]

there are currently two situations where code invariants exist that are both "not great":

• stats modes ansi // ncurses is a compile time option (by editing the makefile in undocumented ways)
• writer compressed/plain (and the lost pcap/pcapng) is an ifelse fest.

i would like to change both into runtime-options. after talking with my personal C guru (C is not my native language), the current rough plan is ...

• start with the stats problem since it is the less complex one.
• add a struct statsmode that has a name and the handler function pointer(s).
• add that struct for all compiled-in stats modes (keep ncurses compile-time optional!) to some stats-modes-avail list via eal init hooks.
• add a rte_log stats mode that does NOT use human readable numbers (because if you are feeding stats to f.ex. grafana, once you hit the TB range for queue bytes, at %.2f that is not really high fidelity stats)
• (very optional) a direct stats feed to carbon/grafana.

going to ask @Woutifier and @groud for review of actual commits, but feedback/suggestions already welcome by all interested parties.

[groud]

stats modes ansi // ncurses is a compile time option (by editing the makefile in undocumented ways)

Ah well, it's not really optional in fact then. But yeah, that can be moved to command line arguments.

writer compressed/plain (and the lost pcap/pcapng) is an ifelse fest.

Not sure I understand, but I am not against improvements there.

The plan looks interesting, but it's hard to understand what you mean without the code. But yeah, structs can be helpful in some situations. Make sure not to over-engineer this too much though, I'd rather avoid having too many abstraction layers for a code that is, for now, easy enough to understand.