Gap-11: Functions Observe (B → A) — Invocation tracing, Durable monitoring, KQL library

Parent: #1608

## Functions Observe: B → A

**Learning from:** App Service (A-) — auto-instrumentation option

**Current state:** Manual SDK instrumentation, ARG queries for monitoring discovery exist. Missing: Functions-specific best practices, invocation-level tracing, Durable Functions monitoring.

## Deliverables

- [ ] `references/functions/observability.md` in `appinsights-instrumentation` — Functions-specific guide:
- FunctionAppLogs table queries for invocation analysis
- Invocation-level tracing (invocation ID, trigger details)
- Durable Functions monitoring (orchestration status, activity timing)
- Host-level vs function-level telemetry
- [ ] Built-in monitoring integration reference (Functions + App Insights auto-connection)
- [ ] KQL query library for common Functions diagnostics (error rate, duration P50/P95/P99, cold start frequency)
- [ ] Add 1 integration test for Functions observability setup

## Key Scenarios

- Invocation duration analysis (P50/P95/P99 by function name)
- Cold start frequency tracking per hosting plan
- Durable Functions orchestration monitoring dashboard
- Scaling event correlation with performance metrics

> ⚠️ **Starting assessment** — domain experts should modify or contribute what they believe is essential.

Start with the proposed references/functions/observability.md and review the existing ARG queries and manual SDK instrumentation described in the issue. Done means the Functions-specific guide, monitoring integration reference, diagnostic KQL library, and one Functions observability integration test are added, covering the listed scenarios.

Written by the indexing model from the issue text.