refactor: Eliminate redundant cleanup logic in entry log extraction

[nodece]

Motivation

The purpose of extractMetaFromEntryLogs is to extract metadata, not perform garbage collection. The logic for removing entry logs with no active ledgers is handled separately in doGcEntryLogs().

Changes

This change removes the redundant cleanup code from extractMetaFromEntryLogs to better separate concerns and avoid duplication.

[StevenLuMT]

rerun failure checks

[nodece]

rerun failure checks

[nodece]

rerun failure checks

[nodece]

rerun failure checks

[AnonHxy]

protectd could be removed. Default access modifiers is enough.

[zymap]

Here is the reason why it did the cleanup: #1124. You could declare a new method to avoid the duplicated code.

[nodece]

@StevenLuMT @zymap

In the current implementation, all metadata is first extracted and stored in a map before performing a full GC pass over the entire dataset. This approach can result in high memory usage and delayed cleanup, especially in environments with large or fragmented metadata stores.

I suggest we should replace the current bulk extraction + GC logic with an incremental streaming approach, as follows:

Current Approach:

Map<String, Metadata> allMetadata = extractAllMetadata();
gc(allMetadata);

Proposed Approach:

for (String key : listAllMetadataKeys()) {
    Metadata metadata = extractMetadata(key);
    gc(metadata);
}

Benefits:

• Reduced memory footprint: Avoids holding all metadata in memory at once.
• More timely cleanup: GC is performed immediately after extraction.
• Improved scalability: Handles large metadata sets more gracefully.

[hangc0276]

The current logic can cleanup those no active ledger's entrylog files immediately.