Add scientific Python library support (numpy, pandas, matplotlib, etc.)

Add numpy, pandas, matplotlib, scipy, seaborn, and statsmodels as
runtime dependencies for numerical methods, statistics, and data
visualisation exercises. Set MPLBACKEND=Agg in the subprocess
environment so matplotlib works headlessly in the Docker container.
Also tightens the Python constraint to >=3.11 to match lf-toolkit.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

Author: m-messer

evaluation_function/evaluation.py

@@ -18,6 +18,7 @@ def _run_code(code: str, stdin: str) -> tuple[str, str, bool]:
             capture_output=True,
             text=True,
             timeout=_TIMEOUT,
+            env={**os.environ, "MPLBACKEND": "Agg"},
         )
         return proc.stdout, proc.stderr, False
     except subprocess.TimeoutExpired: