Update pom.xml to pull MapR's Hadoop version 2.4.1.

[rkannan82]

No description provided.

[AmplabJenkins]

Can one of the admins verify this patch?

[pwendell]

Jenkins, test this please.

[SparkQA]

Test build #25775 has started for PR 4108 at commit 8e1008b.

• This patch merges cleanly.

[SparkQA]

Test build #25775 has finished for PR 4108 at commit 8e1008b.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test PASSed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/25775/
Test PASSed.

[pwendell]

Sure, I can pull it in.

[srowen]

I still am not clear on why this is maintained in Spark (or other vendor distro builds). The result of this profile must be maintained but is not a build you can get support for. The distro already has a presumably fully tested build included for users of the distro. It may be viewed as slight favoritism. It also implies that distros require special builds which is not necessarily true. I understand adding repo references but what is the point of this or distributing distro specific but unsupported builds on the Spark site? Is it that this version is not supported by MapR?