perf: add covering index for airtime chart queries

The airtime/utilization chart queries (get_airtime_data and
get_airtime_buckets) range-scan and order packets by timestamp,
selecting only timestamp/length/payload_length/transmitted. On a large
packets table this forced a full scan of the row heap, saturating slow
storage (e.g. a Pi SD card): each dashboard poll took longer than the
client timeout, aborted polls stacked, and sustained I/O starved the
transmit queue.

Add a covering index on packets(timestamp, length, payload_length,
transmitted) so these queries run index-only, dropping the read from the
full row heap to just the index range. Verified via EXPLAIN QUERY PLAN
(COVERING INDEX idx_packets_airtime). Additive and idempotent.

repeater/data_acquisition/sqlite_handler.py

@@ -191,6 +191,14 @@ class SQLiteHandler:
                 conn.execute(
                     "CREATE INDEX IF NOT EXISTS idx_packets_transmitted ON packets(transmitted)"
                 )
+                # Covering index for the airtime/utilization charts. get_airtime_data
+                # and get_airtime_buckets range-scan and order by timestamp, selecting
+                # only these columns; keeping them all in the index lets SQLite serve
+                # the query index-only, avoiding a full scan of the (large) row heap.
+                conn.execute(
+                    "CREATE INDEX IF NOT EXISTS idx_packets_airtime "
+                    "ON packets(timestamp, length, payload_length, transmitted)"
+                )
                 conn.execute(
                     "CREATE INDEX IF NOT EXISTS idx_adverts_timestamp ON adverts(timestamp)"
                 )