blog: v2 — doubled technical depth, code examples, architecture details

Rewrote blog with significantly more technical content:
- workspace_assembler.py merge logic with actual code
- Tenant ID resolution chain (H2 Proxy → Router → base ID extraction)
- Firecracker microVM 5-phase lifecycle (entrypoint.sh deep dive)
- DynamoDB single-table schema (PK/SK patterns, GSI design)
- S3 directory structure with all paths
- Real-time usage tracking (atomic DynamoDB ADD)
- Plan A constraint injection code
- Plan E regex audit code
- skill_loader.py role filtering with manifest example
- BFS sub-department rollup algorithm
- Memory writeback exclusion rules explained
- 600 lines, 8 screenshots, code-heavy for developer audience
- Charts: placeholder for draw.io replacement (ASCII diagrams inline)

Author: JiaDe