[PERFORMANCE] Presto read Paimon Table very slow

[lfyzjck]

reproduce:

A cdc table

> select * from paimon_table where id = ? limit 1;

Query 20231115_063347_02385_ga6ya, FINISHED, 33 nodes
Splits: 49 total, 49 done (100.00%)
[Latency: client-side: 4:48, server-side: 4:48] [0 rows, 0B] [0 rows/s, 0B/s]

we expected the query finished in 1s but actual we spent about 2 mins. Since parquet/orc footer already contains min/max stats about id, we only need read at most 1 compacted file.

flamegraph: