Fix monitoring performance and reliability

[qaid]

Summary

• Replace per-tick nettop -L 1 (5 s per call, so the "1 s" loop ran every ~6 s and showed speeds ~6x too high) with one persistent nettop -L 0 -s 1 stream; speeds are now delta / measured elapsed time
• Read pipes before waitUntilExit() in the netstat and lsof readers (output > 64 KB deadlocked the monitor loop and froze the app)
• Give Connection a stable id so expanded rows are not rebuilt and DNS lookups do not refire every second
• Run lsof only while the detail panel is visible
• Batch history writes into one SQLite transaction every 10 s (WAL mode, cached DateFormatter, pruneOldData now runs at launch)

Test plan

• swift build passes
• App launches, one nettop child, ~1% CPU, child exits on quit
• Menu bar speed matches Activity Monitor
• App list appears ~5 s after launch and updates each second

🤖 Generated with Claude Code

https://claude.ai/code/session_01G2229uhWSPYZC5KrJNvGBF