Fix discarded return value Scala compiler warnings

[jbaiera]

Subtask of #969

Fix warnings related to discarding non-Unit values from method calls. These should either be checked, or their signatures changed to stop returning things they don't need to.

/Users/`james.baiera/Documents/source/elasticsearch-hadoop/spark/core/main/scala/org/elasticsearch/spark/rdd/EsSpark.scala:102: discarded non-Unit value
    rdd.sparkContext.runJob(rdd, new EsRDDWriter(config.save(), hasMeta).write _)
                           ^
/Users/james.baiera/Documents/source/elasticsearch-hadoop/spark/core/main/scala/org/elasticsearch/spark/rdd/JavaEsRDD.scala:49: discarded non-Unit value
    InitializationUtils.setValueReaderIfNotSet(settings, classOf[JdkValueReader], log)
                                              ^
/Users/james.baiera/Documents/source/elasticsearch-hadoop/spark/core/main/scala/org/elasticsearch/spark/rdd/ScalaEsRDD.scala:51: discarded non-Unit value
    InitializationUtils.setValueReaderIfNotSet(settings, classOf[ScalaValueReader], log)
                                              ^
/Users/james.baiera/Documents/source/elasticsearch-hadoop/spark/core/main/scala/org/elasticsearch/spark/serialization/ScalaValueReader.scala:190: discarded non-Unit value
    map.asInstanceOf[Map[AnyRef, Any]].put(key, value)
                                          ^
/Users/james.baiera/Documents/source/elasticsearch-hadoop/spark/sql-20/src/main/scala/org/elasticsearch/spark/sql/EsSparkSQL.scala:94: discarded non-Unit value
    sparkCtx.runJob(srdd.toDF().rdd, new EsDataFrameWriter(srdd.schema, esCfg.save()).write _)
                   ^
/Users/james.baiera/Documents/source/elasticsearch-hadoop/spark/sql-20/src/main/scala/org/elasticsearch/spark/sql/SchemaUtils.scala:295: discarded non-Unit value
        info._1.setProperty(level, StringUtils.concatenate(fields, StringUtils.DEFAULT_DELIMITER))
                           ^

[takezoe]

Hm, checking discarding return value may be overkill because all of them are defined in Spark or Java libraries. We can't touch them or shouldn't change their signatures. They can be ignored explicitly as:

val _ = rdd.sparkContext.runJob(rdd, new EsRDDWriter(config.save(), hasMeta).write _)

But it's not very common also in Scala world. So eliminating "-Ywarn-value-discard" option might be a solution in this situation.

[jbaiera]

@takezoe I think that makes sense to be honest. This is just a placeholder to clear out these types of warnings for the sake of turning on fatal-warnings. If that ends up as removing the flag because it makes the most sense, I think it makes sense.