Support additional BitNet models

[m96-chan]

Currently only BitNet b1.58 2B-4T (ggml-model-i2_s.gguf) is tested and listed as a preset.

Candidates

• bitnet_b1_58-large (0.7B) — 1bitLLM/bitnet_b1_58-large
  • Previously had BITNET_0_7B_CONFIG; needs GGUF + end-to-end verification
• Falcon3-Edge 1B / 3B — BitNet-quantized Falcon models
• Aramis-2B — Community BitNet model

What "supported" means

1. GGUF file loads without errors
2. Generates coherent text
3. Added as a preset in example apps (if GGUF is publicly hosted)
4. Listed in docs and README

Notes

• Auto-detection from GGUF metadata should handle most models already
• Main blocker is finding reliable public GGUF URLs and verifying output quality
• BITNET_0_7B_CONFIG and BITNET_3B_CONFIG still exist in code but are untested