[SPARK-4423] Improve foreach() documentation to avoid confusion between local- and cluster-mode behavior

Hi all - I've added a writeup on how closures work within Spark to help clarify the general case for this problem and similar problems. I hope this addresses the issue and would love any feedback.

Author: Ilya Ganelin <ilya.ganelin@capitalone.com>

Closes #4696 from ilganeli/SPARK-4423 and squashes the following commits:

c5dc498 [Ilya Ganelin] Fixed typo
07b78e8 [Ilya Ganelin] Updated to fix capitalization
48c1983 [Ilya Ganelin] Updated to fix capitalization and clarify wording
2fd2a07 [Ilya Ganelin] Incoporated a few more minor fixes. Fixed a bug in python code. Added semicolons for java
4772f99 [Ilya Ganelin] Incorporated latest feedback
448bd79 [Ilya Ganelin] Updated some verbage and added section links
5dbbda5 [Ilya Ganelin] Improved some wording
d374d3a [Ilya Ganelin] Merge remote-tracking branch 'upstream/master' into SPARK-4423
2600668 [Ilya Ganelin] Minor edits
c768ab2 [Ilya Ganelin] Updated documentation to add a section on closures. This helps understand confusing behavior of foreach and map functions when attempting to modify variables outside of the scope of an RDD action or transformation