StatusForge 0.6.0 — local release

Local-only release: one Go binary with the embedded web UI, backed by DynamoDB Local on loopback. Nothing is deployed or hosted. Build from source with make build (git describe embeds the version).

• Retention periods with DynamoDB TTL, resumable incident retention and format-6 backfill, permanent deletion
  of archived monitors and applications, private export/import, and a Settings page.
• Single Go binary with the embedded web UI, history fallback, build version in CLI/API/UI, and a demo-data
  banner.
• Loopback Host and mutation Origin/body protections, Content Security Policy, make security-check, a
  reproducible demo, and a bounded capacity report.
• Response-time axis labels switch to seconds at 1 s or more, matching the summary sentence.
• Demo history ends when seeding starts; demo notifications are delivered at their transitions with real HTTP
  statuses.
• A request the browser aborts is logged at INFO with client_canceled=true, not as an ERROR.
• Wide tables scroll from the keyboard as named focusable regions; client-side navigation moves focus to the
  new heading and sets the page title; the date-time picker shows its focus ring.
• make doctor reports Python 3 as optional; local setup gains a first-monitor walkthrough and a note that
  checkouts share local data.
• The scheduler no longer loses most scheduled checks under load. Workers take work as soon as they are
  free and never receive work that has already expired; each scheduling pass lists monitors once and works on
  them with bounded concurrency. On the measured workstation, 100 monitors at 10 s went from 12 completed
  checks in 10 minutes to every slot covered; 100 monitors at 60 s went from 63 gap slots in 5 minutes to 0.
• Scheduler coverage is visible. GET /api/system/scheduler and the Overview report unknown,
  disabled, ok, or behind; the Overview shows Scheduler is behind with guidance; the log adds
  counts to scheduler tick plus scheduler coverage degraded / recovered.
• Known limit: about 10 scheduled checks per second on DynamoDB Local on the measured workstation. At 250
  monitors at 10 s about 39 % of checks ran; the rest were recorded as gaps. See
  docs/operations/capacity-report.md.