Add test coverage for all major components:
- GPT model: architecture, generation, MQA, rotary embeddings (19 tests)
- Inference engine: KV cache, sampling, tool use (17 tests)
- Optimizers: Muon and AdamW functionality (10 tests)
- Checkpoint management: save/load, metadata (5 tests)
- Data loading and utilities (13 tests)
docs: update README with test documentation and learning guide
- Add For Students section with structured learning path
- Document architectural decisions and key concepts
- Add test usage instructions
