test(stats): cover span_days, time bounds, and tie-breaking

[protosphinx]

Why

pm_bench/stats.py has three behaviors that existing tests in test_stats.py do not exercise:

1. span_days, earliest, and latest are only tested on an empty log (where they are trivially 0 / None). No test confirms the calculation on a real log.
2. _top_n_sorted guarantees ties are broken lexicographically, but no test pins that contract.
3. The single-event edge case (span = 0, all length stats = 1) is not covered.

These gaps mean a regression in summarize() could slip through CI undetected. This PR closes them with three targeted unit tests, no production-code changes.

What

• test_summarize_span_days_and_time_bounds: verifies span_days == 2.0 and earliest/latest on the shared _events() fixture (a log that spans exactly 2 days).
• test_summarize_single_event_span_zero: single case, single event; span = 0.0, earliest = latest = the event timestamp, all length stats = 1.
• test_summarize_tie_breaking_is_lexicographic: in _events() both "a" and "b" appear 2 times, and "c"/"x" appear 1 time each; asserts the full ordering [("a",2), ("b",2), ("c",1), ("x",1)].

Tests

• All 3 new tests run via pytest tests/test_stats.py
• Full suite: 251 passing (was 248)

Self-merge gate

• all CI checks pass
• LOC delta < 250 (added + removed)
• no public-API surface change
• no runtime-dependency additions
• no workflow file changes
• tests added or extended

---

Generated by Claude Code