[core] Support merging sparse row ranges for global indexes

[JingsongLi]

Summary

This PR adds opt-in support for building global indexes over sparse, monotonically increasing RowID ranges whose gaps are permanent. It covers BTree global indexes and vector-style single-column global indexes, allowing index metadata to use the enclosing RowID span while reads and writers still honor the actual data ranges.

Changes

• Add btree-index.build.merge-row-ranges and vector-index.build.merge-row-ranges, both disabled by default.
• Preserve actual IndexedSplit row ranges while using the enclosing RowID range for index metadata when sparse ranges are merged.
• Replace the old single-column writer variants with GlobalIndexSingleColumnWriter#write(@Nullable Object key, long relativeRowId) so BTree, vector, Tantivy, and Lumina writers all receive explicit relative RowIDs.
• Wire sparse range merging through core, Flink, and Spark BTree index builders, plus Flink/Spark generic vector index builders.
• Relax global index RowID existence conflict checks only when the corresponding merge option is enabled and the span endpoints are covered by current data files.
• Update vector/full-text test index helpers to persist row ids, so sparse relative RowIDs are returned correctly by test readers.
• Reuse overridable table schemas in the discontinuous RowID BTree test so Lance format coverage keeps using file.format=lance.

Testing

• mvn -pl paimon-core -am -Pfast-build -DfailIfNoTests=false -Dtest=BtreeGlobalIndexTableTest#testBuildBTreeGlobalIndexWithDiscontinuousRowIdRanges test
• mvn -pl paimon-lance -am -Pfast-build -DfailIfNoTests=false -Dtest=LanceBTreeGlobalIndexTest#testBuildBTreeGlobalIndexWithDiscontinuousRowIdRanges test
• mvn -pl paimon-flink/paimon-flink-common -am -Pfast-build -DfailIfNoTests=false -Dtest=GenericIndexTopoBuilderTest test
• mvn -pl paimon-spark/paimon-spark-common -am -Pfast-build -DfailIfNoTests=false -Dtest=CreateGlobalIndexProcedureTest test
• mvn -pl paimon-vector/paimon-vector-index -am -Pfast-build -DfailIfNoTests=false -Dtest=VectorGlobalIndexerFactoryTest,VectorGlobalIndexTest test
• mvn -pl paimon-tantivy/paimon-tantivy-index -am -Pfast-build -DfailIfNoTests=false -Dtest=TantivyFullTextGlobalIndexTest test
• mvn -pl paimon-common,paimon-core,paimon-flink/paimon-flink-common,paimon-spark/paimon-spark-common,paimon-vector/paimon-vector-index,paimon-tantivy/paimon-tantivy-index,paimon-lumina -am -Pfast-build -DskipTests test-compile
• mvn -pl paimon-common,paimon-core,paimon-flink/paimon-flink-common,paimon-spark/paimon-spark-common,paimon-vector/paimon-vector-index,paimon-tantivy/paimon-tantivy-index,paimon-lumina spotless:check
• git diff --check

Notes

These options should only be enabled when RowID gaps are permanent and will not be filled later. With the default values, existing global index build behavior is unchanged.

The local Tantivy test command completed successfully, with Tantivy native test cases skipped by their test assumptions on this machine.