[SPARK-10763][ML][JAVA][TEST] Update Java MLLIB/ML tests to use simplified dataframe construction

[holdenk]

As introduced in https://issues.apache.org/jira/browse/SPARK-10630 we now have an easier way to create dataframes from local Java lists. Lets update the tests to use those.

[srowen]

It seems OK to me. I assume we still have at least one test that tests the version with an RDD argument!

[holdenk]

@srowen So this changes it so that all of the ml tests that created a Dataframe using an RDDs of Rows now use List's of rows as input to the Dataframe constructor, but in the Dataframe constructor tests both are tested.

[SparkQA]

Test build #42915 has finished for PR 8886 at commit 1b45c36.

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

[SparkQA]

Test build #42916 has finished for PR 8886 at commit 4bf4ae6.

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

[mengxr]

fix indentation. (It seems that your format configuration for Java is different from Scala.)

[mengxr]

LGTM except indentation. The test for creating DataFrame from JavaRDD should live in DataFrame's unit tests.

[SparkQA]

Test build #42932 has finished for PR 8886 at commit a668f0d.

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

[mengxr]

Merged into master. Thanks!