chore: exposing max_source_fragments on OPTIMIZE command (#348)

This PR adds support for `max_source_fragments` in the OPTIMIZE command.
This parameter was added to restrict the number of fragments included in
a compaction. It's main use is for incrementally compacting extremely
fragmented datasets.

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: hamersaw

docs/src/operations/ddl/optimize.md

@@ -30,6 +30,7 @@ The `OPTIMIZE` command supports several options to control compaction behavior:
 | `num_threads` | Long | Number of threads for compaction |
 | `batch_size` | Long | Batch size for processing |
 | `defer_index_remap` | Boolean | Whether to defer index remapping |
+| `max_source_fragments` | Long | Maximum number of source fragments to compact in a single task |
 
 ### Examples
 

lance-spark-base_2.12/src/main/scala/org/apache/spark/sql/execution/datasources/v2/OptimizeExec.scala

@@ -49,6 +49,8 @@ case class OptimizeExec(
     argsMap.get("batch_size").map(t => builder.withBatchSize(t.value.asInstanceOf[Long]))
     argsMap.get("defer_index_remap").map(t =>
       builder.withDeferIndexRemap(t.value.asInstanceOf[Boolean]))
+    argsMap.get("max_source_fragments").map(t =>
+      builder.withMaxSourceFragments(t.value.asInstanceOf[Long]))
 
     builder.build()
   }

lance-spark-base_2.12/src/test/java/org/lance/spark/update/BaseOptimizeTest.java

@@ -121,7 +121,8 @@ public void testWithFullArgs() {
                     + "materialize_deletions_threshold=0.2f,"
                     + "num_threads=2,"
                     + "batch_size=2000,"
-                    + "defer_index_remap=true"
+                    + "defer_index_remap=true,"
+                    + "max_source_fragments=128"
                     + ")",
                 fullTable));