fix: When receiving MDN, mark all preceding messages as noticed, even having same timestamp (#7928)

[iequidoo]

This fixes flaky JSON-RPC test_multidevice_sync_seen.
Fix #7928

[iequidoo]

EXPLAIN QUERY PLAN UPDATE msgs SET state=? WHERE state=? AND hidden IN (0,1) AND chat_id=? AND timestamp<? OR state=? AND hidden IN (0,1) AND chat_id=? AND timestamp=? AND id<?;
QUERY PLAN
`--MULTI-INDEX OR
   |--INDEX 1
   |  `--SEARCH msgs USING COVERING INDEX msgs_index7 (state=? AND hidden=? AND chat_id=? AND timestamp<?)
   `--INDEX 2
      `--SEARCH msgs USING COVERING INDEX msgs_index7 (state=? AND hidden=? AND chat_id=? AND timestamp=? AND rowid<?)

Let's prepare for adding timestamp to msgs_index7, it's not the only place where timestamp speeds up a query, e.g. building chatlists is also sped up. See also #7904 for why hidden IN (0,1) (instead of hidden=0) is used in EXPLAIN QUERY PLAN.

[iequidoo]

This can't be rewritten using AND (timestamp<? OR timestamp=? AND id<?), see https://sqlite.org/queryplanner.html#_or_connected_terms_in_the_where_clause -- OR should be the final operator connecting the terms.

[iequidoo]

test_iroh_webxdc.py::test_no_duplicate_messages hung for unknown reason (https://github.com/chatmail/core/actions/runs/23614345419/job/68778915544?pr=8046), but i don't think that's related to the fix here. test_multidevice_sync_seen passed for all configurations