refactor(node/chain): simplify the chain service loop and docs

[tcoratger]

Summary

Clarity and documentation pass over the ChainService that drives consensus timing. No behavior change to the timing logic; the existing chain tests pass (with one test updated for a simplified return contract).

Chain service (chain/service.py)

• Retire the Interval | None sentinel. _initial_tick now returns the anchor interval Interval(0) before genesis instead of None. This collapses the run() loop guard into a single "wait unless a strictly-new interval is due" predicate and removes the post-sleep double-read of the interval. Verified safe: the clock clamps elapsed time to zero before genesis, so total_intervals() never underflows.
• Trim verbose docs. The multi-paragraph _tick_to docstring shrinks to three lines (the skip/yield rationale already lives in the inline comments at the exact lines), stop() becomes one line, and several run() comment blocks lose their narration. All the genuine non-obvious rationale is preserved: the wall clock can step backward, why skipping stale intervals is safe, why we yield and re-read the store, and why catch-up aggregates are discarded.

Sync service (sync/service.py)

• Document a concurrency invariant at the gossip-block store write-back: the processed path never awaits, so a concurrent clock tick cannot advance the store between the read and the write-back. This was verified by exhaustive path analysis (the only suspending paths return processed=False); the comment guards the emergent safety against a future change that makes block processing await.

Testing

• just check passes (ruff, format, ty, codespell, mdformat).
• uv run pytest tests/node/chain tests/node/sync — all pass.

🤖 Generated with Claude Code