[FLINK-3522] [storm compat] PrintSampleStream prints a proper message when involked without arguments

This also includes some minor cleanups

This closes #1689

Temporary work around to restore initial state on failure during recovery as
required by a user. Will be superseded by FLINK-3397 with better handling of
checkpoint and savepoint restoring.

A failure during recovery resulted in restarting a job without its savepoint
state. This temporary work around makes sure that if the savepoint coordinator
ever restored a savepoint and there was no checkpoint after the savepoint,
the savepoint state will be restored again.

This closes #1720.

This closes #1714

* Explicitly state to stop the local environment

This closes #1718

This makes sure that for bounded data sets with watermarks, the final elements get properly reflected in window results.

This closes #1675

This closes #1695

This closes #1713

A second attempt at object swapping in reduce drivers.

This closes #1626

Transitive dependency EvEmitter updated to 1.0.2

This closes #1661

The change improves the display of SerializedListAccumulators in the web
interface by displaying the number of elements contained in the
accumulator instead of the hash code of the object.

This closes #1717.

This closes #1712.

The start computation states have a null event associated. When serializing these states
one has to check whether the event is null or not, because not all serializer can handle
null values.

A single state NFA failed to compute a matching pattern because it was regarded as a
terminal state in the pattern extraction algorithm. By letting the first states start
with a two level dewey number this problem is fixed.

After restart of the execution graph, the test driver waits for the
graph to assign the new executions in a loop. If switchToRunning()
is called multiple times in this loop, the job is canceled. This
could happen, if one execution had its resource assigned and called
switchToRunning(), but the other didn't. In this case, the complete
loop was retried and switchToRunning() was called again.

This closes #1708.

This closes #1706

  - IngestionTime now only auto-generates watermarks
  - EventTime does not auto-generate ingestion watermarks

This also removes the "ExecutionConfig.areTimestampsEnabled()" flag.

This closes #1699

- Move "Windowing" to separate page
  - Make "Working with Time" sub page of "Windowing"
  - Fix link in "Working with Time" page
- Move "Working with State" page to "Fault Tolerance"
  - Add Table of Contents to State Backends page
  - Hierarchical restart strategies sections
- CEP => Event Processing (CEP)

This closes #1705.

This closes #1707.

This closes #1701

This closes #1697

This closes #1581

This changes slot sharing settings to single method
slotSharingGroup(String) on DataStream.

Operations inherit the slot sharing group of the input if all input
operations are in the same slot sharing group.

The default slot sharing group is "default" this can also be explicitly
set using slotSharingGroup("default"). This overrides the inheriting
behaviour.

This closes #1641.

removed unused import

use new Path().toUri()

added missing braces

This closes #1651.

This was causing to many problems with security tokens and yarn. Now,
let the RocksDB backup run in a thread but
don't interrupt these Threads anymore on closing. The Threads will close
themselves because the copy operation will fail because of a
FileNotFoundException when the state directories are being cleaned up.

This also removes the ExternalProcessRunner because it is not needed
anymore and using it causes too many headaches.

This closes #1687.

This closes #1652.

This closes #1688.

Add test case

This closes #1643.

This closes #1666

[FLINK-3438] ExternalProcessRunner fails to detect ClassNotFound exception because of locale settings

[FLINK-3438] Improved solution, no workaround

[FLINK-3438] Change: a faulty process now causes a RuntimeException to be thrown

This closes #1665.