Unexpected performance benchmark exceptions should fail fast

I tested it by starting a second transaction here while one was already associated with the calling thread and then used the commands: cd narayana/ArjunaJTA/jta # after cloning the perf repo mvn clean install -DskipTests java -Xdebug -Xrunjdwp:transport=dt_socket,server=y,suspend=n,address=5005 -classpath target/classes -Xms4096m -Xmx4096m -jar target/benchmarks.jar 'com.arjuna.ats.jta.xa.performance.*StoreBenchmark.*' -t 1 -r 30 -f 0 -wi 5 -i 5 -foe true -rf csv -rff JDBCStoreBenchmark.csv Without the fix this reproducer will cause subsequent executions of the benchmark to log a warning (since one cannot associate a second transaction with a thread unless it is a sub-transaction) and produce very large logs without stacktraces, but with the fix the benchmark will bail out after the first exception with a stacktrace.