Update architecture documentation

This commit significantly expands and restructures the architecture documentation for SWE-AF. Key updates include:

- Renaming the document title to "SWE-AF Architecture" and providing a more detailed overview of its function.
- Incorporating a Mermaid diagram to visualize the high-level workflow.
- Adding a comprehensive Table of Contents for improved navigation.
- Detailing the Build Pipeline across its five phases, including the verify-fix loop and idempotency.
- Elaborating on the Planning → Issue DAG process, including the five planning agents and how issues are decomposed into levels.
- Introducing and explaining core Architectural Patterns:
    - Hierarchical Escalation Control
    - Structured Concurrency with Barrier Synchronization
    - Agent Isolation with Semantic Reconciliation
    - Graceful Degradation with Explicit Incompleteness
    - Runtime Plan Mutation
    - Durable Execution & Checkpoint Recovery
    - Risk-Proportional Resource Allocation
    - Cross-Agent Knowledge Propagation
- Providing a detailed explanation of AgentField Governance, focusing on DID Identity and Execution Verifiable Credentials.
- Including an expanded Agent Catalog that lists agents by role and phase, along with their tools and output schemas.
- Documenting Model Configuration options for agent AI providers.

The documentation now offers a much deeper and more structured understanding of SWE-AF's internal workings, design principles, and operational patterns.

Author: santoshkumarradha