Report errors in tasks to the driver via a Mesos status update

When a task throws an exception, the Spark executor previously just
logged it to a local file on the slave and exited. This commit causes
Spark to also report the exception back to the driver using a Mesos
status update, so the user doesn't have to look through a log file on
the slave.

Here's what the reporting currently looks like:

    # ./run spark.examples.ExceptionHandlingTest master@203.0.113.1:5050
    [...]
    11/10/26 21:04:13 INFO spark.SimpleJob: Lost TID 1 (task 0:1)
    11/10/26 21:04:13 INFO spark.SimpleJob: Loss was due to java.lang.Exception: Testing exception handling
    [...]
    11/10/26 21:04:16 INFO spark.SparkContext: Job finished in 5.988547328 s

1 files changed, 18 insertions, 0 deletions

diff --git a/examples/src/main/scala/spark/examples/ExceptionHandlingTest.scala b/examples/src/main/scala/spark/examples/ExceptionHandlingTest.scala
new file mode 100644
index 0000000000..46f658eab2
--- /dev/null
+++ b/examples/src/main/scala/spark/examples/ExceptionHandlingTest.scala
@@ -0,0 +1,18 @@
+package spark.examples
+
+import spark.SparkContext
+
+object ExceptionHandlingTest {
+  def main(args: Array[String]) {
+    if (args.length == 0) {
+      System.err.println("Usage: ExceptionHandlingTest <host>")
+      System.exit(1)
+    }
+
+    val sc = new SparkContext(args(0), "ExceptionHandlingTest")
+    sc.parallelize(0 until sc.defaultParallelism).foreach { i =>
+      if (Math.random > 0.75)
+        throw new Exception("Testing exception handling")
+    }
+  }
+}