fix(loop): retry empty final LLM responses before fallback

[alexk-dev]

Summary

• add retry logic in DefaultToolLoopSystem for empty final LLM responses (null/blank without tool calls): up to 2 retries (3 attempts total)
• classify persistent empty final responses as structured LLM validation failures and set LLM_ERROR with detailed diagnostics
• keep voice-only responses (voiceRequested / voiceText) out of empty-response retries to avoid false positives
• keep response preparation strict for final-ready empty responses and emit system.error.llm instead of generic fallback
• add fallback reason classification in FeedbackGuaranteeSystem logs for easier production triage
• extend tests in DefaultToolLoopSystemTest and OutgoingResponsePreparationSystemTest to cover retries, recovery, and failure classification

[sonarqubecloud]

Quality Gate passed

Issues
8 New issues
0 Accepted issues

Measures
0 Security Hotspots
84.2% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud