Fix: update _run_script to use individual language per script execution.

[kevtimova]

This PR fixes a bug in the _run_script method where script executions were always returning 0 reward due to incorrect language parameter handling.

Changes:

• Updated _run_script method signature to accept language: str instead of languages: List[str]
• Modified the task creation loop to zip scripts with their corresponding languages
• Updated the sandbox execution call to use the single language parameter

Motivation: The current implementation was incorrectly passing a list of languages to individual script executions, which caused all executions to return 0 reward instead of the expected evaluation scores.