Refactor prompt to treat needs-action as a system label

[AlaeMghirbi]

Why?

The LLM prompt currently performs two distinct tasks in a single request:

1. Action Classification: Determine if email needs action (YES/NO)
2. Label choice: Select relevant labels from user's labels
   => violates the single responsibility principle

Problems identified :

• Two separate task sections with different instructions
• Action is a boolean, labels are IDs

What?

Refactor LlmMailBackendClassifierListener to treat needs-action as a system label rather than a separate classification task.
The goal is to have a single, generic action in the prompt that reads the provided labels (both system and user labels) and then performs the classification.

Before:

OUTPUT: YES,labelA,labelB

After:

OUTPUT: Needs-action,labelA,labelB

How?

1. Update LlmOutput parsing logic:
   Keep only List<String> labels: Parse output as comma-separated label IDs only
2. Change flags' setting logic: No special handling for "action" field
3. Update userPrompt in build_user_prompt()to include the system labels in addition to user context labels
4. Simplify system prompt: Single task: "Select relevant labels"

[coderabbitai]

📝 CodeRabbit Plan Mode

Generate an implementation plan and prompts that you can use with your favorite coding agent.

• Create Plan

Examples

• Example 1
• Example 2

---

🔗 Related PRs

#2058 - ISSUE-2050 LlmMailPrioritizationClassifier: Inject username into the user prompt [merged]

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[chibenwa]

Also mention as a ticket goal to simplify the prompt so that it targets only one task and not 2