feat: add Langfuse telemetry integration

[mhaack]

Summary

• Add src/telemetry.ts — lazy OTEL provider init using BasicTracerProvider + LangfuseExporter, with initTelemetry(env) and flushTelemetry()
• Wire telemetry into streamText via experimental_telemetry in src/server.ts — traces include model, token counts, latency, tool calls, and per-request org/site metadata
• Add langfuse-vercel and @opentelemetry/sdk-trace-base dependencies

How it works

The AI SDK's experimental_telemetry: { isEnabled: true } routes spans through the global @opentelemetry/api tracer. initTelemetry(env) registers a BasicTracerProvider + LangfuseExporter on first request per Worker isolate. If LANGFUSE_PUBLIC_KEY or LANGFUSE_SECRET_KEY are not set, telemetry is silently skipped — chat is unaffected.

Test Plan

• 31 tests passing (npm test)
• Type-check clean (npm run type-check)
• Set LANGFUSE_PUBLIC_KEY and LANGFUSE_SECRET_KEY via wrangler secret put (production) or .dev.vars (local)
• Send a chat request and verify a trace appears in the Langfuse dashboard with model name, token counts, and org/site metadata

[github-actions]

🎉 This PR is included in version 1.4.0 🎉

The release is available on:

• v1.4.0
• GitHub release

Your semantic-release bot 📦🚀