Fix clock thread spinning at midnight due to missing tm_mday increment

[jamesmulcahy]

Summary

• publish_time_and_date never incremented tm_mday when crossing midnight, so mktime normalised the struct back to midnight of the current day (already in the past) — causing sleep_until to return immediately and the date message to be published twice in quick succession each night before recovering
• The existing manual rollover check (tm_mday >= 32) was unreachable dead code (tm_mday was never incremented) and would have been wrong anyway for February and 30-day months
• Fix: increment tm_mday and remove the manual rollover logic; mktime is C-standard-specified to normalise out-of-range struct tm fields (e.g. January 32 → February 1, month 12 → January of next year) including leap-year-aware February, and is already called on the very next line

Test plan

• Deploy and observe clock MQTT messages across a midnight boundary — expect exactly one publish per minute with no double-fire at 00:00
• Verify correct date rollover at end of month (including February in a non-leap year) and end of year

🤖 Generated with Claude Code

[tpanajott]

Thanks for this. I've spent a bit of time on this issues a few times but never quite got it to work as I wanted. Need to verify that it works as expected at midnight but should be fine as far as I understand it.