Add detailed model card with training details

Update chat template to olmo_thinker (with <think> generation prompt)

Upload OLMo-3 7B checkpoint (NeoX -> HF conversion, global_step62)

initial commit