IMPALA-4266: Java udf returning string can give incorrect results

The memory management of string results was wrong: strings returned from
Exprs must live until the next time FreeLocalAllocations() is called.
Otherwise the buffer holding the string is freed or reused by the next
UDF call. The fix is to copy string values into a buffer with the
right lifetime.

Testing:
Added a regression test based on Bharath's example that reproduced the
bug reliably.

Change-Id: I705d271814cb1143f67d8a12f4fd87bab7a8e161
Reviewed-on: http://gerrit.cloudera.org:8080/4941
Reviewed-by: Tim Armstrong <tarmstrong@cloudera.com>
Tested-by: Internal Jenkins

testdata/workloads/functional-query/queries/QueryTest/java-udf.test

@@ -290,3 +290,26 @@ INT
 999
 999
 ====
+---- QUERY
+drop table if exists replace_string_input
+====
+---- QUERY
+create table replace_string_input as
+values('toast'), ('scone'), ('stuff'), ('sssss'), ('yes'), ('scone'), ('stuff');
+====
+---- QUERY
+# Regression test for IMPALA-4266: memory management bugs with output strings from
+# Java UDFS, exposed by using the UDF as a grouping key in an aggregation.
+# The UDF replaces "s" with "ss" in the strings.
+select distinct udf_test.replace_string(_c0) as es
+from replace_string_input
+order by 1;
+---- TYPES
+string
+---- RESULTS
+'sscone'
+'ssssssssss'
+'sstuff'
+'toasst'
+'yess'
+====