This commit changes how the rpcConnection is managed in JobManagerLeaderListener under leader change.
This component clears now the fields rpcConnection and currentJobMasterId if the leader loses leadership.
Moreover, it only restarts a connection attempt if the leader session id is new.

This closes #11603.

Before, when using PackagedProgramUtils (for example in the standalone
cluster entrypoint or the web ui) the Flink Configuration would not be
applied to the execution environment.

This also adds a test that verifies that we forward configuration.

This closes #11594

This closes #11376

This closes  #11569 .

This closes #11527

This PR introduces alternatives for methods that are using the String based expressions in TableEnvironments.

This PR also removes all usages of the String based expressions methods. It is a preparation for deprecating and in result removing those methods in the future.

The motivation is to help debugging FLINK-15772.

Before, we were checking for an exact number of restarts, this is too
strict, because we could have more than the expected number of restarts.
Now we check whether we have gte the number of expected restarts.

[FLINK-16674][python][docs] Add documentation about how to use user-defined metrics for Python UDF (#11576)

With inverting the dependency between flink-clients and flink-streaming-java,
it is now necessary to explicitly add the flink-clients dependency in order
to execute the walkthrough jobs from within the IDE.

This closes #11551.

Flink quickstarts' poms need to include flink-clients as a dependency because it is
no longer transitively pulled into the project.

This commit adds a flink-clients dependency to all example modules. That
way they become executable from the IDE again.

This closes #11572

jquery is used for the documentation.

This closes #11559

This closes #11580

This closes #11575

Clearing the rpcConnection field in the JobManagerLeaderListener when target loses leadership
prevents that we try to reconnect to the target in case JobLeaderService.reconnect(JobID) is
called.

This closes #11552.

This commit fixes a bug where we did not update the state of the AkkaRpcActor
in case of terminating it. Moreover, this commit fixes the problem that the
onStop action could have been called multiple times. Last but not least, it
changes the enum names of the state implementations for better diagnostics.

This closes #11549.

In order to not fail the job submission with a TimeoutException if
it takes longer than 10 s, this commit increases the web.timeout to
10 minutes.

This closes #11565.

Because the classloader is now closed when the task fails the UDF only has access to the bootstrap classloader, which doesn't contain our own test classes.

Allows user classloader can be unloaded even if a reference on the context classloader outlives the user code.

The test is actually testing FlinkUserCodeClassLoaders.

This closes #11504

This closes #11560

* Some join order changes for blink-planner due to the rule fire sequence changes, see https://github.com/apache/calcite/commit/35caa059a762094c7df0b30e9b51358a19b48ac2, they are still correct
* The Correlate row count estimation has been fixed from a always 1 to join like estimation, thus, if the inputs of Join is a Correlate, the join algorithm would very probably changes, i.e. batch.sql.SubplanReuseTest
* Due to CALCITE-3729, the filter condition was pushed down for some Join cases: batch.sql.join.JoinReorderTest
* Due to CALCITE-2450 RexNode normalization, the predicates sequence of some test changes: logical.subquery.FlinkRewriteSubQueryRuleTest
* The Decimal modulus precision inference has been fixed: planner.expressions.DecimalTypeTest

* GROUP_ID translation was fixed