Update prompt for scope 3 category emissions to return an array with detailed emissions to improve quality and reliability

[Greenheart]

Update how garbo is extracting scope 3 emissions:

• Update prompt with detailed instructions: For each scope 3 category, return an array with all detailed emissions. For example for scope 3 category 6 business travel, return an array of all detailed emissions, like [{ "label": "flights", "value": 2531 }, {"label": "trains", "value": 12 }]. If no detailed emissions are found for that scope 3 category, return an empty array, like []. If there's only a number summarizing emissions in that category, return the number instead.
• Use OpenAI structured output to ensure we get output in the correct format.
• Use our code to calculate the total emissions for each scope 3 category, and don't let the LLM do maths. This will improve quality and reliability.
• In the future, consider saving the detailed emissions for each scope 3 category, to allow more valuable comparisons between companies and industries.