Add configurable proactive low-memory warning notification before killing

[benedictjohannes]

Problem

The current -n / --notify option in earlyoom only sends a D-Bus notification after a process has already been SIGTERM’d or SIGKILL’d.
On desktop systems this is too late — users have no chance to react (close tabs, save work, kill the real culprit themselves) before the automated killer kicks in.

This has been discussed in #54 but issue was closed probably due to earlyoom not supporting notification at that time. Since earlyoom has notification now, this becomes a viable/possible alternative.

Proposed solution

Add an optional early warning notification that fires noticeably before the kill thresholds are reached.

New flags (both optional, preserve 100 % backward compatibility):

Flag	Meaning	Suggested default
-w MULT, --warn-mult=MULT	Warning threshold = kill threshold × MULT (in %)	none (disabled)
--warn-backoff=SEC	Minimum seconds between repeated warnings	300

Example usage

# Normal kill thresholds
earlyoom -m 10,5 -s 80,50

# Add warning 25 % earlier than the kill threshold + backoff
earlyoom -m 10,5 -s 80,50 -w 125 --warn-backoff 300

→ Memory warning at 12.5 % available (instead of 10 %), swap warning at ≈ 62.5 % used, etc.

Implementation impact

• Uses the exact same memavail/swaptotal calculations earlyoom already does every poll cycle
• Hooks into the existing main loop
• No extra polling, no new threads, no new syscalls
• Uses the exact same notification code path as the existing -n (just a different summary/body, e.g. “Warning: Low memory – consider closing applications”)

This would turn earlyoom into a friendly heads-up system.

Happy to provide a draft PR if you like the idea.

Thanks for maintaining the best userspace OOM killer out there!