spec(logging): add Ralph Loop logging specification

Define detailed logging requirements for observability of model decision-
making and tool execution. Includes scenarios for:
- Session context logging (iteration start, model, prompt source)
- Tool execution logging (start events, results with timing/status, error summaries)
- Model reasoning logging (intent changes, decision explanations)
- Fitness evaluation logging (command progress, score breakdown)
- GitHub API logging (issue creation, comment posting, retry attempts)
- Model rotation logging (stall detection, escalation to premium models)
- Error and warning prominence (errors with context, timeouts)
- State persistence logging (checkpoints, resume info)
- Consistent log format with timestamp, level, message
- Multiline indentation support for readability
- Optional log level filtering (DEBUG/INFO/WARN/ERROR)
- Quiet mode via RALPH_QUIET env var

Author: Addono