[SPARK-6196] [BUILD] Remove MapR profiles in favor of hadoop-provided

[srowen]

Follow up to #7047

@pwendell mentioned that MapR should use hadoop-provided now, and indeed the new build script does not produce mapr3/mapr4 artifacts anymore. Hence the action seems to be to remove the profiles, which are now not used.

CC @trystanleftwich

[SparkQA]

Test build #41319 timed out for PR 8338 at commit 14bcab4 after a configured wait of 175m.

[SparkQA]

Test build #1675 timed out for PR 8338 at commit 14bcab4 after a configured wait of 175m.

[rxin]

I will let @pwendell respond to this when he comes back from vacation.

[srowen]

@rxin yeah that's fine, there's no emergency, although these profiles are already unused; the release script for 1.5 will not produce artifacts that use these. On the other PR Patrick indicated, I think, that he intends these to be removed in favor of hadoop-provided

[pwendell]

Yes this LGTM - these are outdated and I don't even think MapR is advising their customers to use these. They are asking people to use hadoop-provided, which was created to simplify using Spark with different MapR versions.

[srowen]

Nice, I do agree with that, and think it will be a better experience for all. The vendor <repository> declarations can stay as a convenience. I'd be inclined to do something similar for the cdh4 profile. I can see that it functions as a sort of generic hadoop-2.0.0 build but would not mind zapping it.