Update to Mesos 0.12.1

Print run command to stderr rather than stdout

Properly account for killed tasks.

The TaskState class's isFinished() method didn't return true for
KILLED tasks, which means some resources are never reclaimed
for tasks that are killed. This also made it inconsistent with the
isFinished() method used by CoarseMesosSchedulerBackend.

Minor clean-up in metrics servlet code

Add MetricsServlet for Spark metrics system

Simple PageRank algorithm implementation in Python for SPARK-760

Implementing SPARK-865: Add the equivalent of ADD_JARS to PySpark

Now ADD_FILES uses a comma as file name separator.

Create SparkContext in beforeAll for MLLib tests

Change scala.Option to Guava Optional in Java APIs

Clean up scaladoc in ML Lib.

Changed yarn.version to 2.0.5 in pom.xml

Remove now dead code inside of docs

Also build and copy ML Lib scaladoc in Spark docs build.
Some more minor cleanup with respect to naming, test locations etc.

Report compressed bytes read when calculating TaskMetrics

This overcomes test failures that occur using Maven

Use new Configuration() instead of slower new JobConf() in SerializableWritable

Refactor SGD options into a new class.

Fixed path to JavaALS.java and JavaKMeans.java, fixed hadoop2-yarn profi...

JobConf's constructor loads default config files in some verisons of
Hadoop, which is quite slow, and we only need the Configuration object
to pass the correct ClassLoader.

Adding Scala version of PageRank example