Add diagnostic logging for debugging eval runs

[udid-aws]

Problem

When skill-eval functional, trigger, or report produces unexpected results, there's no way to see what's happening internally:

• Which Claude command was invoked
• What working directory was used
• What Claude returned (stdout/stderr)
• How trigger signals were classified
• What files were copied into temp workspaces

This makes it difficult to diagnose issues like incorrect AWS profiles, trigger detection gaps, or workspace misconfiguration.

Proposed Solution

Add global --debug and --debug-log FILE flags that emit DEBUG-level logs from the key modules (agent_runner, functional, trigger) using Python's stdlib logging — zero external dependencies.

Related

PR: #8