Rewrite legacy model card for public users

Clarify legacy Gemma 4 incompatibilities

Mark model card as legacy

Add README with base model metadata

Copy config.json from google/gemma-4-31B-it

Upload tokenizer - SAE introspection LoRA - google/gemma-4-31B-it-layers_15-30-45-decoder-True_latentqa_cls_past_lens_gemma-4-31B-it - final model

SAE introspection LoRA - google/gemma-4-31B-it-layers_15-30-45-decoder-True_latentqa_cls_past_lens_gemma-4-31B-it - final model

initial commit