Enhance SchemaQuake documentation and training notebook. Updated Blog.md and README.md to clarify the RL environment aspect of SchemaQuake. Added supporting evidence for the training results in training-results.md. Improved the training notebook with clearer instructions and code for model loading and evaluation. Adjusted demo app to reflect changes in model loading and checkpoint handling.

Load trained model repo in demos.

Add GPU adapter demo paths.

Enhance evaluation metrics and documentation for imperfect heuristic agent.

Rename mini-blog to Blog.