feat: `meshd doctor` — expand health checks (inotify, stale devices, conflict pile)

[gitcoder89431]

Summary

Expand meshd doctor with more actionable health checks beyond basic connectivity.

Motivation

The current doctor command checks if Syncthing is reachable and the folder is registered. There are several common failure modes that silently degrade the experience that we can detect and explain.

Proposed Checks

Linux / Arch specific

✔ inotify limit: 524288 (recommended ≥ 131072)
✘ inotify limit: 8192  — too low, Syncthing may miss file changes
  Fix: echo fs.inotify.max_user_watches=524288 | sudo tee /etc/sysctl.d/40-meshd.conf && sudo sysctl -p

Stale / ghost devices

⚠ MacBook-Old: not seen in 47 days — consider removing (meshd peer rm <id>)

Conflict pile

⚠ 12 unresolved conflicts — run `meshd conflicts --fix` to resolve

Versioning disk usage

⚠ .stversions is 4.2 GB — consider pruning old versions
  Oldest version: notes.md (143 days ago)

Folder out of sync

✘ ~/meshd not fully synced to Phone: 73% (23 files remaining)

API key validity

✔ Syncthing API key: valid
✘ Syncthing API key: expired or changed — re-run meshd to auto-detect

CLI Output Format

meshd doctor           # human-readable, colored pass/warn/fail
meshd doctor --json    # machine-readable for scripts / MCP tool

Implementation Notes

• inotify limit: read /proc/sys/fs/inotify/max_user_watches on Linux only, skip silently on macOS
• Stale devices: compare lastSeen from /rest/stats/device against a 30-day threshold
• Conflict pile: walk ~/meshd for files matching .sync-conflict-*
• Versioning size: du on ~/meshd/.stversions
• All checks should be non-fatal — doctor always exits 0 unless --strict flag is passed

[gitcoder89431]

Doctor expanded: inotify watch count (Linux), conflict pile check, .stversions disk usage, per-device sync completion %, stale device warnings.