detect netdata exit reasons

[ktsaou]

Exit reasons

• track the reason of exit, including signals, api quit command, cmd exit command, windows service stop.
• detects a netdata update by comparing the existing netdata binary file timestamp and size to the one found during exiting
• detects system shutdown my monitoring for systemd dbus events for PrepareForShutdown events (and PrepareForSleep - which we don't do anything with currently)

Status file

• a status file now keeps on disk information about the running agent, including the first fatal message encountered and its stack trace (with libunwind).
• the file is JSON and is serialized back, when netdata restarts
• crash detection is now based on the new status file
• interrupted shutdown is now based on the new status file

Runtime ephmemeral files

• Netdata uses /run/netdata (or an equivalent on all operating systems) for all ephemeral runtime files (.pid, .pipe, .lock)
• Netdata uses file lock to ensure it does not start multiple times.

JSON/logfmt generation and parsing

• Netdata now generates json that \n, \r, \t, \f and \b are printed like this, instead of \uXXXX.
• Logfmt strings are also escaping these sequences the same way.
• systemd-cat-native now defaults in treating \n as newline characters (it was disabled by default).

External plugins logging to journal

• fixed a bug that prevented external plugins log to native journal format.

High-Performance tuning

• use more replication threads
• spread tier1+ points collection over time, to remove sudden spikes in data collection
• eliminate most BUFFER rellocs for normal data collection and replication
• use 4 libc arenas on parents