[spark] Correct MigrateTableProcedureTest with set target table iterations

[ArnavBalyan]

Purpose

• MigrateTable UT is flaky and causes CI failures.
• The UT runs 3 migrate $format for (parquet/orc/avro) with the same hive_tbl_$random and target_$random (same random shared at class level).
• Run 2 re-uses cached HMS values from SessionCatalog.tableRelationCache from run 1, which is a parquet file, and appends .orc to it. (making it an invalid .parquet.orc when the entries are cached).
• This causes the UT to fail, ensure we can use different names to avoid cache collision

Tests

• UT now passes

[ArnavBalyan]

cc @JingsongLi @leaves12138 could you pls take a look when possible thanks!