feat(clock): persist software clock to NVS on boards without an RTC

[parsazolfaghar]

What

Persist the software clock to NVS on boards without an RTC chip, and restore it on boot.

Why

Boards like the T-Deck have no RTC, so a time set via manual/NTP/GPS is lost on a full
power-off and the clock resets to the build default on the next boot. This periodically
saves the current epoch to NVS and restores it at startup, so the clock comes back close
to correct even without WiFi/GPS at boot time.

Addresses part of #1866.

How

• Adds restore_persisted_time() + time_persist_task() in main.cpp, guarded by
  #if !defined(HAS_RTC) so RTC builds are completely unaffected.
• Stored in NVS (via Preferences), not the user filesystem, so it survives FS
  reformats and custom/non-standard partition layouts.
• First save fires within ~60s of the clock being set, then at most every 5 min
  (bounds NVS wear; worst-case staleness ~5 min).
• On boot it restores the saved epoch (with a post-2025-01-01 sanity check) before the
  rest of setup().

Behavior / limitations

Without an RTC, elapsed time during a power-off cannot be tracked, so the clock restores
the last-known time and will be behind by however long the device was off; it re-syncs
to exact time via NTP/GPS when available. This is a best-effort "don't reset to the default
date" fallback, not a hardware-RTC replacement.

Testing

Built and tested on a LilyGO T-Deck Plus (lilygo-t-deck-pro): set the time, fully
power-cycled, and the clock restored the last-known time instead of resetting to the build
default. Confirmed it survives a custom/dual-boot partition layout (NVS-backed). A 30-minute
power-off returned ~30 minutes behind, as expected for a no-RTC software clock. RTC builds
are unaffected (compiled out via !defined(HAS_RTC)).

[bmorcelli]

This is indeed a good idea, although it loses time sync every reboot, but will at least keep date info, important in file metadata..

Rebase it to dev brench and add these new funtions changes into the ones in src/core/utils.* as clock functions reside there and intefrate with the updateClockTimezone() if needed 😉

[Ninja-jr]

This is indeed a good idea, although it loses time sync every reboot, but will at least keep date info, important in file metadata..

Rebase it to dev brench and add these new funtions changes into the ones in src/core/utils.* as clock functions reside there and intefrate with the updateClockTimezone() if needed 😉

Or...just an idea, but if it's about the reboot time making it slightly lose sync, one could calculate the average reboot time (of course it depends on devices and if quick boot is enabled or not) and add that when re-syncing time. If quick boot option on then add just the quick boot time.

Of course still not perfect but it would probably be more accurate.

This probably won't solve the problem on a prolonged power off still, but that one would be hard to solve anyway. Wouldn't need a full power off, but rather a deep sleep with periodic rts sync...but that would be a big battery drainer anyway.

[parsazolfaghar]

Thanks for the review! Done both:

• Rebased onto dev (PR retargeted).
• Moved the functions into src/core/utils.* alongside the clock functions — restorePersistedClock() + the save task now live there and are called once from init_clock() on the non-RTC path, overriding the default date with the last-saved epoch. Guarded by !defined(HAS_RTC) so RTC builds are unaffected.

Let me know if you'd like any further tweaks.