refactor: ensure compatibility with the latest Langfuse version

[a-klos]

This pull request introduces significant updates to the RAG infrastructure and evaluation codebase, primarily focused on improving Langfuse integration, updating dependencies, and refactoring configuration for external service management. The changes enhance traceability, simplify environment setup, and ensure compatibility with the latest Langfuse version.

Langfuse Integration and Evaluation Refactor:

• Refactored the Langfuse evaluation workflow in langfuse_ragas_evaluator.py to use context managers (start_as_current_generation and run) for better trace and generation handling, and replaced manual scoring and linking with direct trace scoring. [1] [2]
• Updated the dataset creation method to use the new Langfuse API signature (create_dataset(name=...)).

Dependency Updates:

• Upgraded the langfuse Python package from 3.0.0 to 3.3.4 across all relevant pyproject.toml files for improved compatibility and features. [1] [2] [3]

Infrastructure Configuration Improvements:

• Updated the Langfuse Helm chart dependency version from 0.12.1 to 1.5.1 in Chart.yaml.
• Overhauled values.yaml to externalize configuration for PostgreSQL, Redis (KeyDB), ClickHouse, and S3/MinIO, removing embedded environment variable definitions and introducing dedicated configuration sections for each service. [1] [2] [3]
• Removed internal deployment of MinIO and Valkey, and clarified that KeyDB is used for Redis functionality. [1] [2] [3]

Tracing Enhancements:

• Improved tracing in traced_runnable.py by updating span names to reflect the actual chain class and including input/output in the trace, enabling more detailed observability.