Create an env for turn_left policy

[NotAnyMike]

• Create env
• Get the exact init params
• Modify reset
• Random angle
• Figure out how to find if it is inside the first track (shape) (use ray casting algorithm)
• Modify step: check every step if agent is still in correct path, else rest
• Check why max_time_out not working
• Implement it using the same environment for everything
• Test it