fix(extractor): streaming with guardrails + catalyst_type normalization

- Switch Ollama calls from non-streaming to streaming with early termination
- Add loop detection, max token limit, and stall timeout guards
- Add catalyst_type alias normalizer to handle model hallucinations
- Add explicit enum values in extraction prompt for catalyst_type
- Add streaming config knobs to OllamaConfig

services/extractor/prompts.py

@@ -114,6 +114,8 @@ Fill these fields:
 
 For each company entry fill: ticker, company_name, relevance (0-1), sentiment, impact_score (0-1), impact_horizon, catalyst_type, key_facts (list), risks (list), evidence_spans (verbatim quotes from text).
 
+catalyst_type MUST be exactly one of: earnings, product, legal, macro, supply_chain, m_and_a, rating_change, other. Use "other" if none of the specific categories fit.
+
 --- DOCUMENT TEXT ---
 {document_text}
 --- END DOCUMENT TEXT ---"""