Overview
This PR updates DatabaseMetaData.getIndexInfo() to address an issue where calling ResultSet.next() after exhaustion could throw an invalid cursor position exception.
The fix removes the use of CachedRowSet for combining sp_statistics and columnstore index results, and instead uses a UNION ALL query. This ensures standard JDBC cursor behavior while keeping columnstore index support intact.

Problem Description
The current implementation uses CachedRowSet to merge results from sp_statistics and Columnstore index queries. However, CachedRowSet has stricter cursor validation than SQLServerResultSet:

if (cursorPos < 0 || cursorPos >= numRows + 1) {
    throw new SQLException("Invalid cursor position");
}

This causes SQLException: Invalid cursor position when next() is called after the ResultSet has been fully enumerated, which differs from standard JDBC behavior where such calls should simply return false.

Root Cause
Columnstore index support was first added in #2598
to address missing indexes in sp_statistics results. That implementation relied on CachedRowSet to merge sp_statistics output with additional Columnstore index query results, which in turn caused the cursor positioning issue.

Solution

• SQL Server: Replaced CachedRowSet merging with a single UNION ALL query (INDEX_INFO_COMBINED_QUERY) that combines sp_statistics results with Columnstore index data from sys.indexes.
• Azure Synapse DW: Execute sp_statistics separately and combine with sys.indexes via UNION ALL ensuring comprehensive index coverage including columnstore indexes.

This eliminates intermediate result set handling and ensures consistent behavior across environments.

Testing

• Validated cursor behavior on both SQL Server and Azure Synapse DW.
• Added comprehensive JUnit tests for ResultSet navigation patterns.
• Verified Columnstore indexes are properly included.
• Confirmed sp_statistics compatibility maintained.

Closes #2758 #2778