[improve] Increase the cache prefetch success rate when entryId is discontinuous.

[zhuhai1221]

bookkeeper/bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/storage/ldb/SingleDirectoryDbLedgerStorage.java

Line 674 in b660726

ByteBuf entry = entryLogger.readEntry(originalLedgerId,

When Pulsar's persistence is configured with EnsembleSize != WriteQuorum, for example, EnsembleSize=2 and WriteQuorum=1, entry prefetching fails. If there is a backlog of messages, after the RocksDB index cache is evicted, the disk read rate becomes very slow, and entry prefetching fails because the entryId is validated, and at this point, the entryId is discontinuous.

Is it possible to read entries without validating the entryId?

use
org.apache.bookkeeper.bookie.storage.EntryLogger#readEntry(long)