fix(core): implement robust auto-discovery for model layers

fix(core): correct layer paths for HRM and Qwen

chore: remove diagnostic logs

chore: switch default model to Qwen/Qwen2.5-0.5B-Instruct

debug(core): add model structure logging

fix(core): add HRM-Text-1B layer path and attention attribute

Switch default model to HRM-Text-1B

Deploy Thought Anchors