Fix sample rate handling and improve HTML report

[solomon23]

Summary

• Fix 1Hz assumption bug: The spike detector assumed 1 sample/second, but the O2Ring S records at 4-second intervals. This caused recording lengths to show ~1/4 of actual time and inflated all per-hour metrics (SI/hr, TAB, PC10/hr, etc.) by ~4x. Fixed by resampling to 1Hz in load_data() when CSV timestamps indicate non-1s intervals. Works with any sample rate.
• Fix chart IndexError: Chart generation used original CSV row count for time axis but resampled event indices. Now generates a 1Hz time axis matching the resampled data.
• Report improvements: Add year to dates, add editable Notes column from filename remarks, remove redundant Night Label column, auto-uncheck daytime sessions, skip detector_results.csv from analysis, add explicit white background for dark mode viewers.

Screenshot

Test plan

• Verify recording lengths match filename durations (e.g., _8h_26m.csv should show ~8h 26m)
• Verify per-hour metrics are reasonable (SI/hr ~5-30 for typical nights, not ~20-120)
• Verify charts generate without IndexError
• Verify daytime sessions (start 6am-8pm) are unchecked by default
• Verify Notes column shows filename remarks
• Test with 1Hz data (original O2Ring) to confirm no regression

🤖 Generated with Claude Code

[AJolly]

Thanks for yourr changes! can you check against the latest version? And your bot is wrong - the o2ring-s works on 1second resolution, not 4. The older o2ring works on 4 second resolution

Hope the project is useful for you