Reindex subset of vertices

[ntisseyre]

Summary

This PR introduces a significant optimization to the reindexing process in JanusGraph by allowing a subset of vertices to be reindexed instead of scanning the entire storage.
This enhancement provides substantial performance improvements, primarily when the specific subset of vertices for indexing is already known.

NOTE

This feature is currently supported only for CQL storage. Other storage backends still need to be implemented.

KeyColumnValueStore.java

KeyIterator getKeys(final List<StaticBuffer> keys, final SliceQuery query, final StoreTransaction txh) throws BackendException {
        throw new NotImplementedException();
    }

Motivation

Previously, reindexing required scanning all vertices in storage, which could be highly resource-intensive and time-consuming, particularly in large datasets.
This update enables users to focus on a targeted subset of vertices, reducing the time and computational load for reindexing. This is especially beneficial in environments where only specific vertices are relevant to a given index or data update.

Changes

• Added the ability to specify a subset of vertices to include in the reindexing process.
• Optimized the indexing engine to skip unnecessary vertices, focusing only on those specified in the subset.

API in JanusGraphManagement

  /**
     * Updates the provided index according to the given {@link SchemaAction} for
     * the given subset of vertices.
     *
     * @param index
     * @param updateAction
     * @param vertexOnly Set of vertexIds that only should be considered for index update
     * @return a future that completes when the index action is done
     */
ScanJobFuture updateIndex(Index index, SchemaAction updateAction, List<Object> vertexOnly);

Benefits

• Improved Performance: By narrowing down the scope of vertices, the reindexing process is much faster and more efficient.
• Resource Optimization: Reduces CPU and memory usage during reindexing by avoiding a full scan.
  Enhanced Flexibility: This feature allows users to update specific sections of the graph more easily without impacting the entire dataset.

Backward Compatibility

This feature is backward compatible and does not impact existing functionality. Users not specifying a subset will still experience the previous behavior of scanning the entire storage.

[porunov]

Thank you @ntisseyre !
Looks great! I have just two small comments.

[porunov]

LGTM. Thank you @ntisseyre !