feat: enrichers offline split_sentences + extract_iocs_text

Para probar la app sin depender de red (DDG bloquea con captcha desde
ciertas IPs). Ambos aplican grouping (umbral 50, preview K=10) replicando
el patron de web_search.

- split_sentences: parte texto en frases (regex), crea nodos Sentence
  conectados con SENTENCE_OF.
- extract_iocs_text: variante de extract_text_entities que lee directo
  metadata.text/description/name, sin requerir fetch previo. Vendoriza
  extract_iocs_py_cybersecurity. Multi-tipo, agrupado en un solo Group
  heterogeneo (decision 6 multi-grupo-por-tipo es fase 2).
- Tipo Sentence en types.yaml.

Tests pytest cubren below/above threshold para ambos.

enrichers/extract_iocs_text/manifest.yaml

@@ -0,0 +1,11 @@
+id: extract_iocs_text
+name: "Extract IoCs from text"
+description: "Extrae IoCs (IPs, emails, dominios, hashes, crypto wallets, CVEs, MAC, telefonos) directamente del texto del nodo. No requiere fetch previo. Sin red."
+applies_to: [text, Text]
+emits: [Email, IPAddress, Domain, FileHash, CryptoWallet, CVE, MACAddress, Phone]
+relations: [EXTRACTED_FROM]
+uses_functions:
+  - extract_iocs_py_cybersecurity
+params:
+  - { name: types, type: string, default: "", description: "CSV de tipos a extraer; vacio = todos" }
+  - { name: max_entities, type: int, default: 500 }