[SPARK-30067][CORE] Fix fragment offset comparison in getBlockHosts

### What changes were proposed in this pull request?

A bug fixed about the code in getBlockHosts() function. In the case "The fragment ends at a position within this block", the end of fragment should be before the end of block，where the "end of block" means `b.getOffset + b.getLength`，not `b.getLength`.

### Why are the changes needed?

When comparing the fragment end and the block end，we should use fragment's `offset + length`，and then compare to the block's `b.getOffset + b.getLength`, not the block's length.

### Does this PR introduce any user-facing change?

No.

### How was this patch tested?
No test.

Closes apache#26650 from mdianjun/fix-getBlockHosts.

Authored-by: madianjun <[email protected]>
Signed-off-by: Dongjoon Hyun <[email protected]>

Author: madianjun

sql/core/src/main/scala/org/apache/spark/sql/execution/PartitionedFileUtil.scala

@@ -64,13 +64,14 @@ object PartitionedFileUtil {
       offset: Long,
       length: Long): Array[String] = {
     val candidates = blockLocations.map {
-      // The fragment starts from a position within this block
+      // The fragment starts from a position within this block. It handles the case where the
+      // fragment is fully contained in the block.
       case b if b.getOffset <= offset && offset < b.getOffset + b.getLength =>
         b.getHosts -> (b.getOffset + b.getLength - offset).min(length)
 
       // The fragment ends at a position within this block
-      case b if offset <= b.getOffset && offset + length < b.getLength =>
-        b.getHosts -> (offset + length - b.getOffset).min(length)
+      case b if b.getOffset < offset + length && offset + length < b.getOffset + b.getLength =>
+        b.getHosts -> (offset + length - b.getOffset)
 
       // The fragment fully contains this block
       case b if offset <= b.getOffset && b.getOffset + b.getLength <= offset + length =>