[SPARK-12025] [SparkR] Rename some window rank function names for SparkR

[yanboliang]

Change cumeDist -> cume_dist, denseRank -> dense_rank, percentRank -> percent_rank, rowNumber -> row_number at SparkR side.
There are two reasons that we should make this change:

• We should follow the naming convention rule of R
• Spark DataFrame has deprecated the old convention (such as cumeDist) and will remove it in Spark 2.0.

It's better to fix this issue before 1.6 release, otherwise we will make breaking API change.
cc @shivaram @sun-rui

[SparkQA]

Test build #46815 has finished for PR 10016 at commit c56a7bc.

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

[sun-rui]

code change looks good. New names have no conflict with R base package.

But I don't see they are deprecated now? Could you point to me the discussion or JIRA that they will be removed in Spark 2.0?
If they are deprecated, I am fine with the name change.
If they are not deprecated, could we add new names as aliases, while keeping the old names at the same time?

[yanboliang]

@sun-rui Thanks for your comments. You can get the issue from the source code and this PR(#9930).

[sun-rui]

@yanboliang, thanks for your info. I didn't look at the latest code:)

LGTM

[shivaram]

LGTM. Thanks @yanboliang -- BTW were these APIs were present in SparkR 1.5 ? If so we'll need to update the release docs about this change.

[felixcheung]

@shivaram they were added only for Spark 1.6, so no need to update release doc on breaking changes
dc3220c
40c77fb

But the new names seem to mask methods from dplyr.