Fix DeploymentHealthCheck: pre-swap staging validation, post-swap warmup wait, and automatic rollback

[Copilot]

The DeploymentHealthCheck subagent blindly swapped staging → production on any regression, with no guard on staging health and no warmup window before measuring results — causing a case where post-swap latency hit 436ms vs a 26ms baseline.

Root causes

• No pre-swap staging check — agent swapped regardless of whether staging was actually faster; if staging was also degraded, the swap worsened production
• 84-second post-swap measurement window — slot swaps trigger cold starts that spike latency for 3+ minutes, making the immediate re-check meaningless
• No rollback path — a swap that didn't help stayed in production indefinitely

Changes

SubAgents/DeploymentHealthCheck.yaml

• Step 7 – Pre-swap staging check: queries staging App Insights before acting; skips swap if staging response time is unavailable or ≥ production (SwapExecuted=false)
• Step 9 – Warm-up wait: calls WaitInMilliSeconds(180000) after swap before any post-swap measurement
• Step 10 – Post-swap verification + auto-rollback: re-queries App Insights; if still >20% over baseline, immediately swaps back and sets outcome "Swap reverted - post-swap still degraded"
• Extended pre-swap query window ago(2m) → ago(5m) for more stable averages
• Added WaitInMilliSeconds to tools list
• Enriched GitHub issue template and Teams post to include staging metrics, swap timestamp, and outcome label

README.md

• Updated demo flow description and ASCII diagram to reflect the new 3-phase remediation: staging check → swap + warmup → post-swap verify → optional rollback