chore: remove GGUFs (moved to lthn/LEM-Gemma3-1B-GGUF)

docs: link to GGUF repo

docs: match 4B card format — quick start first, full quant table, collection links

feat: add full GGUF range (3-bit to 16-bit, 12 quants) + updated model card

feat: LEM-Gemma3-1B with all formats (safetensors, GGUF bf16, GGUF Q4_K_M)

Initial upload: LEM-Gemma3-1B (Ethics-Composure-Ethics sandwich, CL-BPL teacher model)

initial commit