Fix silent message rejection when companion RTC diverges from app clock

[wbijen]

Fix silent message rejection when companion RTC diverges from app clock

Problem

When a companion device's RTC runs ahead of the app's clock, messages sent to Room Servers (and Repeaters/Sensors) are silently rejected. This happens because last_timestamp mixes two different clock sources:

1. Login (ANON_REQ) uses the companion's RTC → e.g. last_timestamp = 1500
2. Messages (TXT_MSG) use the app's clock → e.g. sender_timestamp = 1001
3. The replay check sender_timestamp >= last_timestamp fails: 1001 >= 1500 → rejected silently

The user sees a successful login but their messages never arrive. No error is shown on either side.

This is the same root cause described in #1551 — the proposed fix there was to change the companion side, but that broke repeater counting for channel messages. This PR fixes it on the server side instead, which avoids touching the companion radio at all.

Solution

Don't store the login timestamp in last_timestamp. Keep last_timestamp purely for message/request replay protection (where the clock source is consistent).

Login replay protection is already handled by hasSeen() in the mesh layer, which deduplicates packets by their SHA256 hash. The existing replay check at login time (sender_timestamp <= client->last_timestamp) still works — it just compares against the last message timestamp rather than the last login timestamp.

The fix is applied to all three server types:

• simple_room_server
• simple_repeater
• simple_sensor

Walkthrough

Step	Before (broken)	After (fixed)
Login (RTC=1500)	last_timestamp = 1500	last_timestamp unchanged (stays 0)
Message (app=1001)	1001 >= 1500 → rejected	1001 >= 0 → passes
Message (app=1002)	never reaches here	1002 >= 1001 → passes
Re-login (RTC=1600)	1600 > 1500 → passes	1600 > 1002 → passes

Testing

• Login from companion with RTC ahead of app clock → messages should no longer be silently dropped
• Login replay protection still works via mesh-layer hasSeen() deduplication
• Message replay protection is unaffected (still uses last_timestamp from previous messages)

[usrflo]

Hello @wbijen, applying this fix allowed me to login on the first attempt but failed on subsequent logins.
So I proposed #2834 to separate the login from the message replay timestamps and this works in my tests.