FleetLens 0.2.0

Fixed

• Disk WARNING and CRITICAL results were never reported: the collector emitted
  statuses with trailing whitespace that the classifier did not match.
• FLEETLENS_EMAIL_SEND_ON_OK/WARNING/CRITICAL were ignored; run and email
  now honour them.
• Hosts whose collection aborted early were silently dropped from the report;
  they now appear as CRITICAL with no result collected.
• A single failed or unreachable host made ansible-playbook exit non-zero and
  aborted the whole run before rendering. Rendering now continues, and a stale
  raw report from a previous run is never reused.
• Errors supplied by the collector did not affect host status.
• Per-host status was lost when a JSON report was reloaded for email.
• Heartbeat pings could raise on timeouts or malformed URLs.
• Unreachable hosts reported reboot_required: false instead of unknown.

Added

• A non-zero exit code from the apt, systemd, or journal check on a reachable
  host is reported as a WARNING (<check> check failed (rc=N)).
• needrestart results are included in JSON and shown in Markdown details
  (informational; they do not change status).
• fleetlens run --fail-on none|warning|critical for scheduler-visible exit codes.
• fleetlens email --force to send regardless of status toggles.
• fleetlens console script entry point.

Changed

• scripts/run-check.sh is a thin wrapper around fleetlens run, so cron and
  systemd timer runs behave identically (heartbeats, email gating).
• Memory "used" excludes page cache and buffers.

Removed

• The container-era build prompt (prompts/MAIN.md).