feat: upgrade hook system with multi-executor types, structured output, and event telemetry

[yishuiliunian]

Summary

• Upgrade hook system from shell-command-only to full Harness Hook architecture with Command/HTTP/Prompt executor types, structured output interpretation, and condition-based matching with glob expressions
• Wire all 8 hook events (PreToolUse, PostToolUse, PreRequest, PostInput, Stop, SessionStart, SessionEnd, PreCompact) with correct data flow
• Add configurable agent control-loop thresholds (HarnessConfig), per-turn metrics (TurnMetrics/TurnCompleted), and event causality tracking (event_id/turn_id/correlation_id)

Changes

New files (19):

• loopal-hooks: executor.rs, executor_command.rs, executor_http.rs, executor_prompt.rs, output.rs, input.rs, service.rs, async_store.rs + 3 test files
• loopal-config: harness.rs, hook_condition.rs, resolver_hooks_test.rs
• loopal-kernel: hook_factory.rs
• loopal-protocol: event_id.rs, event_payload.rs
• loopal-runtime: turn_metrics.rs, fire_hooks.rs

Modified files (60):

• Hook config types: HookEvent 8 variants, HookType enum, HookConfig new fields, cross-layer dedup
• Kernel: hook_registry → hook_service, DefaultExecutorFactory with graceful None handling
• Runtime: tool_pipeline pre/post hook rewrite, 8 event wiring points, rewake_rx 3-way select
• Protocol: AgentEvent gains event_id/turn_id/correlation_id, TurnCompleted variant
• Auto-mode: configurable thresholds for LoopDetector/CircuitBreaker/AutoClassifier
• All AgentEvent/HookConfig/AgentLoopParams construction sites updated

Test plan

• bazel build //... passes
• bazel test //... — 48/48 tests pass
• bazel build //... --config=clippy — zero warnings
• Condition glob matching integration tests (registry_test.rs)
• Executor factory validation tests (None provider, missing URL)
• Hook dedup test (resolver_hooks_test.rs)
• CI passes