Fix Hadoop 2.2.0 support, upgrade to Spark 0.8.1

[jey]

This is a continuation of @arahuja's PR #39 with the following changes:

1. Fixed the "No Filesystem found for scheme: ..." errors by merging the META-INF/services files when building the shaded assembly JAR.
2. Modifies the POM so that the default hadoop.version setting of 2.2.0 produces a working build without having to additionally enable a Maven profile.

[AmplabJenkins]

All automated tests passed.
Refer to this link for build results: https://amplab.cs.berkeley.edu/jenkins/job/ADAM-prb/35/

[AmplabJenkins]

All automated tests passed.
Refer to this link for build results: https://amplab.cs.berkeley.edu/jenkins/job/ADAM-prb/37/

[massie]

Can you confirm that Hadoop 1.0.4 still works?

When I followed the instructions in the comment and tried to build for 1.0.4 I got the following error:

Cause: java.lang.ClassNotFoundException: scala.util.Try$

which is caused I believe from using Scala 2.9.2 instead of 2.9.3.

[arahuja]

Matt that is the issue that was occurring under the main (no profile) build with my change. I looked into that and looks like that can be resolved by explicitly pulling in the correct scala version. The issue came up in the Spark project as well, see: mesos/spark#688

[jey]

Good catch. I added an explicit dependency on scala-library and it then built against Hadoop 1.0.4. I'm now testing to see whether 2.2.0 still works as expected.

[AmplabJenkins]

All automated tests passed.
Refer to this link for build results: https://amplab.cs.berkeley.edu/jenkins/job/ADAM-prb/38/

[jey]

I guess Jenkins beat me to the punch.

[massie]

Thanks, Jey!