Add structured logging to core modules

[oscarvalenzuelab]

Add structured logging to core modules

Summary

The dispatcher, pipelines, and poller currently have zero logging imports. Diagnostic output relies on print / rich which is fine for the CLI surface but useless for production observability — when a launchd service silently misbehaves, there's nothing to grep.

Add Python logging to the core modules so operators can debug without re-running with breakpoints.

Motivation

• Enterprise users evaluating the project see "no logging" as a red flag.
• When dev-sync runs as a launchd / systemd daemon, print calls land in the StdOut log but with no structure (no timestamps, no level, no module).
• Several recent issues (fix(dispatcher): resolve bare claude binary name at construction time #16 dispatcher PATH bug, Orchestrator leaks worktrees and branches when dispatcher fails before claude runs #17 cleanup-on-failure) would have been faster to diagnose with a real log trail.

Acceptance criteria

• src/dev_sync/core/dispatcher.py uses logging.getLogger(__name__) and logs:
  • INFO at session start (session_id, repo, issue, branch)
  • INFO at each checkpoint state transition
  • WARNING when claude binary resolution falls back
  • ERROR on subprocess failure with exit code and stderr tail
• src/dev_sync/core/poller.py (or equivalent) logs:
  • INFO when a poll cycle starts/ends
  • INFO when a new issue is claimed
  • WARNING when an issue is skipped (already claimed, missing config, etc.)
• src/dev_sync/pipelines/dev_pipeline.py and secops_pipeline.py log major lifecycle events at INFO and unexpected states at WARNING/ERROR.
• CLI entry point configures a sane default handler:
  • Default level INFO; --verbose / -v flag bumps to DEBUG.
  • Format includes timestamp, level, module name.
  • Honors DEV_SYNC_LOG_LEVEL env var if set.
• Existing user-facing print / rich.print calls in the CLI layer are left alone — this issue is about library logging, not CLI output.
• No log message contains secrets (tokens, repo paths beyond what's already public, etc.).

Out of scope

• Replacing rich CLI output with logs.
• JSON-formatted logs (can be a follow-up if a user asks).
• Remote log shipping.

Notes

• Do not assign this issue.
• Touches src/, so after merge: restart the launchd / systemd services (per docs/operations.md) so they pick up the new code.