[SPARK-4719][API] Consolidate various narrow dep RDD classes with MapPartitionsRDD

[rxin]

MappedRDD, MappedValuesRDD, FlatMappedValuesRDD, FilteredRDD, GlommedRDD, FlatMappedRDD are not necessary. They can be implemented trivially using MapPartitionsRDD.

[SparkQA]

Test build #24086 has started for PR 3578 at commit eb1a89b.

• This patch merges cleanly.

[SparkQA]

Test build #24086 has finished for PR 3578 at commit eb1a89b.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[AmplabJenkins]

Test FAILed.
Refer to this link for build results (access rights to CI server needed):
https://amplab.cs.berkeley.edu/jenkins//job/SparkPullRequestBuilder/24086/
Test FAILed.

[SparkQA]

Test build #24112 has started for PR 3578 at commit eed9853.

• This patch merges cleanly.

[SparkQA]

Test build #24112 has finished for PR 3578 at commit eed9853.

• 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/24112/
Test PASSed.

[pwendell]

LGTM - the main value of these classes is that when you are making a slide to describe RDD transformations it's nice to put "MappedRDD" "FilteredRDD" etc. While useful pedagogically, that's hardly a reason to keep them around :P

[rxin]

Yea - I was thinking maybe we can change the toString implementation for MapPartitionedRDD and add a field similar to the old one. We can do that later if there is demand.

[rxin]

Merging in master.