feat: add graceful degradation helpers and no-op fallbacks

[imran-siddique]

Summary

Every toolkit consumer implements the same try/except import pattern for graceful degradation when toolkit packages are optional:

try:
    from agent_os_kernel import PolicyEvaluator
    _TOOLKIT_AVAILABLE = True
except ImportError:
    _TOOLKIT_AVAILABLE = False

This pattern appears in 5+ files in ai-native-team daemon (trust.py, reliability.py, authority.py, governance.py, conversation_guardian.py).

Proposal

The toolkit should provide:

• governance.available() -> bool — check if toolkit is installed
• No-op fallback implementations that pass-through without enforcement
• A single import: from agent_governance_toolkit import get_policy_evaluator (returns real or no-op)
• Optional dependency marker so consumers can pip install agent-governance-toolkit[full] or agent-governance-toolkit[minimal]

Why This Matters

Graceful degradation is essential for tools that work both with and without the full governance stack. But each consumer reimplements identical try/except blocks, leading to inconsistent fallback behavior and boilerplate.

Real-World Evidence

ai-native-team daemon has this exact pattern in 5 files, each with slightly different fallback behavior. A toolkit-native solution would standardize this.