[CH] Use max_block_size controller read mergetree block size (#3669)

What changes were proposed in this pull request?
Use spark.gluten.sql.columnar.backend.ch.runtime_settings.max_block_size controller read mergetree bolck size. Defalut is DEFAULT_BLOCK_SIZE=65409

(Fixes: #3667)

How was this patch tested?
Test by UT.

cpp-ch/local-engine/Parser/SerializedPlanParser.cpp

@@ -367,7 +367,7 @@ DB::QueryPlanPtr SerializedPlanParser::parseMergeTreeTable(const substrait::Read
         query_context.storage_snapshot,
         *query_info,
         context,
-        4096 * 2,
+        context->getSettingsRef().max_block_size,
         1);
     QueryPlanPtr query = std::make_unique<QueryPlan>();
     steps.emplace_back(read_step.get());