feat(backend): implement M1-M2 core runtime, isolated supervisors, and SQLite persistence

[BerryUIKI]

Description

This PR delivers the foundational backend core runtime, execution drivers, model inventory, and local SQLite persistence for Berry AI Studio (Milestones 1 and 2).

Summary of Changes:

• Deterministic Caching & DAG Engine (\�ackend/app/core/): NodeHash deterministic hashing, cycle detection, and ancestor resolution.
• Durable Storage (\�ackend/app/storage/): SQLite database schema, atomic project persistence, content-addressable asset storage, and multi-root model catalog with pure-Python .safetensors\ header inspection.
• Isolated Engine Supervisors (\�ackend/app/runtime/): ComfyUI and SD WebUI supervisors without host Python fallback, isolated virtualenv installer with interruption recovery, hardware detection, and secure credential storage with key masking.
• Execution Drivers (\�ackend/app/runners/): High-level creative runner for txt2img, img2img, inpaint, and upscale; ComfyUI macro graph compiler; WebUI REST driver; and cloud BYOK driver.
• API Endpoints & Automated Tests (\�ackend/app/main.py, \�ackend/tests/): Implemented endpoints and added test suites (\ est_m1_core.py\ through \ est_m6_launcher_manager.py).

Verification:

• All 68 automated pytest tests pass locally in \�ackend/tests\ (100% passing).