[SPARK-10786][SQL] Take the whole statement to generate the CommandProcessor

In the now implementation of `SparkSQLCLIDriver.scala`:
`val proc: CommandProcessor = CommandProcessorFactory.get(Array(tokens(0)), hconf)`
`CommandProcessorFactory` only take the first token of the statement, and this will be hard to diff the statement `delete jar xxx` and `delete from xxx`.
So maybe it's better to take the whole statement into the `CommandProcessorFactory`.

And in [HiveCommand](https://github.com/SaintBacchus/hive/blob/master/ql/src/java/org/apache/hadoop/hive/ql/processors/HiveCommand.java#L76), it already special handing these two statement.
```java
if(command.length > 1 && "from".equalsIgnoreCase(command[1])) {
  //special handling for SQL "delete from <table> where..."
  return null;
}
```

Author: huangzhaowei <carlmartinmax@gmail.com>

Closes #8895 from SaintBacchus/SPARK-10786.

sql/hive-thriftserver/src/main/scala/org/apache/spark/sql/hive/thriftserver/SparkSQLCLIDriver.scala

@@ -290,7 +290,7 @@ private[hive] class SparkSQLCLIDriver extends CliDriver with Logging {
     } else {
       var ret = 0
       val hconf = conf.asInstanceOf[HiveConf]
-      val proc: CommandProcessor = CommandProcessorFactory.get(Array(tokens(0)), hconf)
+      val proc: CommandProcessor = CommandProcessorFactory.get(tokens, hconf)
 
       if (proc != null) {
         // scalastyle:off println