Add comprehensive documentation for integrating Kagenti with Langfuse cloud
service for LLM observability and trace collection.

## What's Added

- docs/langfuse-integration.md: Complete integration guide including:
  - Langfuse cloud account setup instructions
  - API key generation and configuration
  - OTEL collector configuration for trace export
  - Agent instrumentation with GenAI semantic conventions
  - Testing and verification procedures
  - Cost tracking explanation
  - Troubleshooting guide

- deployments/envs/langfuse-cloud-values.yaml.example: Helm values template
  for configuring OTEL collector to export traces to Langfuse cloud

- README.md: Updated observability section to reference Langfuse (fixed typo:
  Langflow -> Langfuse) and added LLM Observability documentation links

- .gitignore: Added langfuse-cloud-values.yaml to prevent committing API keys

## Why Langfuse Cloud vs Self-Hosted

Attempts to deploy Langfuse v3 self-hosted on Mac M1 with Kind/Rancher failed
due to ClickHouse compatibility issues with nested virtualization. ClickHouse
hangs during startup in the Mac M1 -> Podman/Lima -> Kind -> ClickHouse stack.

Langfuse cloud provides:
- No infrastructure management (no ClickHouse, Redis, PostgreSQL, MinIO, Worker)
- Works from any cluster (Kind, cloud, OpenShift)
- Free tier available (Hobby plan, no credit card)
- Production-ready with automatic scaling and HA

## Cleanup

Removed temporary investigation/research markdown files that were used during
debugging but should not be in the main repository:
- ARCHITECTURE_EXPLAINED.md
- CLICKHOUSE_*.md (3 files)
- END_TO_END_SUCCESS.md
- LANGFUSE_*.md (4 files)
- PODMAN_COMPOSE_SUCCESS.md

These were interim documentation files created during troubleshooting and are
not needed for the final integration documentation.

Resolves investigation into LLM observability options for Kagenti on Mac M1
development environments.

Signed-off-by: Alan Cha <[email protected]>