First cleanup attempt, mostly on Scala code, to follow guidelines from http://docs.scala-lang.org/style:
1. Remove return statement from Scala code where it is not necessary (end of method definition)
2. Remove extra ; from Scala and Java code
3. First drop to abide to 100 chars per line for Scala code.
   Will send another PR for other files as I have encountered.
4. Remove parentheses for empty argument methods that do not have side effect
   (see http://www.artima.com/pins1ed/composition-and-inheritance.html#i1343251059-1)
5. Remove unused import statements in Scala code as I have encountered them.

This is first drop to refactor and clean up Scala code to see comment/ response from community.
More PRs for follow up code will come.

Close #64

This closes #54

This closes #51

This closes #56

This closes #29

I adjusted the error message.

Author: Markus Holzemer <markus.holzemer@gmx.de>

Closes #50 from markus-h/iteration_translation_exception and squashes the following commits:

34a6f87 [Markus Holzemer] adjusted error message when encountering a translation failure for iterations

Taskmanagers are now displayed in the web interface @rmetzger

Author: Jonathan <Jonathan@Hasenburg.de>

Closes #43 from JonathanH5/webclient and squashes the following commits:

5cc5f72 [Jonathan] Finished Flink-928 and simplified how new pages to the WebInterface are added

So people won't encounter and ask the same question as I did just now ;)

Author: Raymond Liu <raymond.liu@intel.com>

Closes #49 from colorant/doc and squashes the following commits:

90bf2aa [Raymond Liu] Mention that Oracle JDK 6 library will fail to build

Add stubs for other internals pages

[ci skip]

Update text on internals overview.

[ci skip]

[ci skip]

[ci skip]

- In NepheleMiniCluster the memory size was divided twice by the number of TaskManagers
- In HardwareDescriptionFactory the memory fraction config value was multiplied too early

This adds support in the TypeExtractor and also one can now use field
expressions for grouping, join, and co-group.

This is required because we need state to detect recursive data types.
This will become necessary when introducing Pojo Types.

https://issues.apache.org/jira/browse/FLINK-978

Author: Jonathan <Jonathan@Hasenburg.de>

Closes #46 from JonathanH5/fixWebI and squashes the following commits:

13a7d41 [Jonathan] Fixed Bug Flink-978

[ci skip]

Change string construction to work around a known JVM performance bug still present in some older JVM versions.

[FLINK-971] Configure PooledByteBufAllocator in NettyConnectionManager instead of using the default allocator

Configuration:
    - 0 heap arenas,
    - n direct arenas (where n = num incoming + num outgoing network IO threads), and
    - bufferSize << 1 bytes page size.

Additionally, OutboundEnvelopeEncoder directly implements ChannelOutboundHandlerAdapter instead of the
MessageToByteEncoder<Envelope> wrapper to have tighter control of memory allocations.

This closes #38.

This closes #39.

This closes #33.

This closes #28.

This closes #17.

[FLINK-973] [FLINK-969] Unify, clean up, and extend all environment logging at JobManager/TaskManager startup

This closes #40.