Add Qwen3.5 local models and remove sub-1B models

[xAlcahest]

Summary

• Add Qwen3.5 family to local model registry (bartowski GGUF repos, download URLs verified):
  • Qwen3.5 9B Q4_K_M (5.5GB), marked as recommended
  • Qwen3.5 4B Q4_K_M (2.7GB)
  • Qwen3.5 2B Q4_K_M (1.3GB)
• Remove three sub-1B models too small for usable post-processing:
  • Qwen3 0.6B
  • Qwen2.5 0.5B
  • Qwen3.5 0.8B
• i18n description keys added across all 9 supported languages

Closes #423