Adds message passing between Hadoop's async resource manager client and
the YarnJobManager actor.

This closes #1561.

This is needed for streaming library features, is identical to the batch API.

Closes #1480

Henceforth, we shall only use the term lower in the context of timestamps
to denote that an event has occurred before another event.

This closes #1534

This closes #1551

Before, the user-specified StateBackedn would not be set when generating the
JobGraph if checkpointing was disabled.

This closes #1516

This closes #1552

This closes #1542

This keeps the log cleaner in case of failed timers while canceling tasks.

Several packages had been updated locally by users when compiling the templates
without changing the version in bower.json. The following packages have been
explicitly updated.

 jquery              2.1.4  →   2.2.0
 angular            1.3.15  →   1.4.8
 angular-moment      0.9.2  →  0.10.3
 angular-ui-router  0.2.13  →  0.2.15
 bootstrap           3.3.5  →   3.3.6
 d3                  3.5.5  →  3.5.12
 dagre-d3           0.4.10  →  0.4.11
 font-awesome        4.3.0  →   4.5.0

This closes #1525

-parallelism is stored Value object within the OperationInfo, so it can be passed as a reference to multiple operations (in cases where a set is internally executed as multiple operations)
-setParallelism is called for every DataSet with either a user-set value or env.getParallelism
-added a DataSink set, providing access to name() and setParallelism() for sinks

This closes #1465.

-identifier saved in java OpInfo
-changed default values to prevent null exceptions
-all operations use the same routine to transfer parameters
-PyPlRcv can handle Tuple0
-labeled py OpInfo fields as transferred/internal
-fixed broadcast OpInfo not having correct identifier
-removed unused projection code

[FLINK-3258] [runtime, streaming-java, tests] Move registerInputOutput code to invoke and remove registerInputOutput

This closes #1538

Before, it could happen that a Task is canceled during snapshotting.
Some State Backends would silently swallow exceptions resulting from
this and the Task would get stuck until the cleanup logic gets to it.

Now, we rethrow a CancelTaskException if isRunning is false in
StreamTask after performing snapshots.

This also moves the logic that swallows exceptions in case a task is not
running anymore from StreamTask to the async caller in Task.

- add suffixes
- adapt change-scala-version script

This closes #1529.

this commit support full outer join includes:
    1. left outer join with REPARTITION_HASH_FIRST.
    2. right outer join with REPARTITION_HASH_SECOND.
    3. fullouter join with REPARTITION_HASH_FIRST and REPARTITION_HASH_SECOND.

this close #1469

This closes #1539.

This closes #1550

This closes #1548

This closes #1522

- Unify usage of expression keys: FLINK-1666
- Unify supported key expressions (incl. support for partitioning on atomic types: FLINK-3233)
- Remove removal of duplicate keys: FLINK-1903
- Unify checks for sort keys
- Add more tests for SelectorFunctionKeys and ExpressionKeys

This closes #1520

This closes #1549.

This closes #1541

Only keep Time for specifying time durations/intervals.

This closes #1512

This closes #1547

This closes #1531

This closes #1530

This closes #1526

Before this commit, Kryo runs extra logic to track and resolve repeated references to
the same object (similar as JavaSerialization)

This disables reference tracking because
  - reference tracking is costly
  - it is virtually always unnecessary in the datatypes used in Flink
  - it is inconsistent with Flink's own serialization (which does not do reference tracking)
  - it may have problems if elements are read in a different order than they are written.

This closes #1528

[FLINK-3246] Remove unnecessary '-parent' suffix from projects 'flink-contrib' and 'flink-streaming-connectors'.

This closes #1515