Fix core dumps triggered by rocksdb compacting when shutdown bk

[AnonHxy]

Descriptions of the changes in this PR:

Fix #4705

Motivation

Fix #4705

Changes

1.Setting the compacting flag of entryLocationIndex as true when shutdown LedgerStorage to stopping the subsequent compact.
2. When LedgerStorage shutdown we will wait unit the compact end.

[lhotari]

LGTM

[lhotari]

There's a need to do some changes in KeyValueStorageRocksDB class too.
Usually RocksDB triggers compaction on it's own without explicit compaction and that type of compaction would continue to cause problems.

Claude suggested this type of pattern:

try {
    // Signal compactions to stop
    db.cancelAllBackgroundWork(true);
} catch (RocksDBException e) {
    // log but continue — we still want to close
} finally {
    // Always close column family handles first, then the DB
    for (ColumnFamilyHandle cfh : columnFamilyHandles) {
        cfh.close();
    }
    db.close();
}

It might be sufficient to simply have this logic in KeyValueStorageRocksDB.close

[lhotari]

a minor detail is that columnFamilyDescriptors aren't closed. I'd assume that it leaks a tiny bit of memory in cases where the JVM process wouldn't terminate (most likely only in test scenarios). I'm not exactly sure about this. It's not a blocker for this PR.

[lhotari]

It's documented here: https://github.com/facebook/rocksdb/wiki/rocksjava-basics#opening-a-database-with-column-families . Not very great docs, but based on that, columnFamilyDescriptors should be closed after the db is closed.

[lhotari]

LGTM

[lhotari]

wrapping this call in a separate try-catch would be recommended

[lhotari]

nit: escaped -> elapsed