* Rename ECS Hook and Operator

* Added Set Skipped button to tasks

* Cleanup

* PR changes; wrong function layout; tested

* Cleanup

* Update more typing-friendly enum
Co-authored-by: Tzu-ping Chung <uranusjr@gmail.com>

* Import for TaskInstanceState and tested locally
Co-authored-by: Tzu-ping Chung <uranusjr@gmail.com>

Previously we pinned this version as v12 as a change to Kube library
internals meant v1.Pod objects now have a logger object inside them, and
couldn't be pickled on Python 3.6.

To fix that we have "backported" the change in Python 3.7 to make Logger
objects be pickled "by name". (In Python 3.7 the change adds
`__reduce__` methods on to the Logger and RootLogger objects, but here
we achieve it `copyreg` stdlib module so we don't monkeypatch
anything.)

This fix is also applied in to airflow core in a separate commit, but we
also apply it here in the provider so that cncf.kubernetes client
library can be updated but still used with older versions of Airflow
that don't have this fix in.

If the user code adds a log handler to write to the standard out/err, there
will be an infinite loop of logging messages. The `StreamLogWriter._buffer` will
be doubled in each iteration, and soon use up all the memory.

This change doesn't stop the infinite loop; it merely changes the growth of
`StreamLogWriter._buffer` from exponential to linear, which will also lead to
runtime error (stack overflow). However, that is much easier to debug (see the
infinite loop) than the out-of-memory error (no call stack).

I think that this doc could be improved by adding examples of how to reference the params in your dag. (Also, the current example code causes this: #20559.)

While trying to find the right place to work a few reference examples in, I ended up rewriting quite a lot of it.
Let me know if you think that this is an improvement.

I haven't yet figured out how to build this and view it locally, and I'd want to do that as a sanity check before merging it, but I figured get feedback on what I've written before I do that.

closes: #20545
Fixed docs for time and random macros as the reference to what they are was incorrect.

Add GoogleCalendarHook to integrate with Google Calendar.

The #20624 broke PRs that are changing setup.py.

This PR fixes it.

The constraints generation was only happening in push/scheduled
runs, but sometimes it is useful to check what constraints would
be generated even in the PRs that change setup.py/setup.cfg

The change causes constraint generation also in the PRs and only
pushing the updated constraints is not executed in PRs.

The test on a busy system could be flaky - the second call
could have not fired.

But this is perfectly ok as we only check the first call which will
always happen.

The script for generating issue for "Provider status" and release
process did not work well when only subset of providers were released.
The issue was generated including some already released packages
even if they were not released in recent batch of providers (if there
was not even a doc change since last release, the package was considered
as being released again).

This PR fixes it by adding a flag that only considers packages that
are present in dist folder (which matches the process of release
manager)

The process has also been updated with more accurate description of
the steps to take - including manual execution of the script rather
than using Breeze (Breeze is not neede for this script).

After removing Python 3.6, some of the excludes have been
empty - which caused error when evaluating combinations to run
im main or when FULL_TESTS were needed.

This PR brings sane excludes

Twine (which we use to upload packages to PyPI) has the
ability to run checks of packages before uploading them.

This allows to detect cases like when we are using forbidden
directives in README.rst (which delayed slightly preparing the
December 2021 provider packages and resulted in #20614

With this PR Twine check will be run for all packages in CI
before we even attempt to merge such change that could break
them.

The dask_executor test was failing occasionally on busy systems
and seems that it was caused by being throttled by other tests.

The 30 seconds timeout in the test seems to be too short. Changed
it to 120 seconds by default (it's just a timeout so it does
not really impact speed of execution of the tests but it gives
the test extra time to complete in case it is throttled.

Timeout was added to assert method so that we can control it
individually in different tests.

* Elasticsearch uses rather 9200 than 80 as default.

* Fix MyPy issues in ``airflow/macros``

**Before**:
```
root@57b2ac1779ad:/opt/airflow# mypy --namespace-packages airflow/macros
airflow/macros/__init__.py:83: error: Argument 1 to "diff_for_humans" of "DateTime" has incompatible type "Optional[datetime]"; expected "Optional[DateTime]"
        return pendulum.instance(dt).diff_for_humans(since)
                                                     ^
Found 1 error in 1 file (checked 2 source files)
```

**After**:
```
root@57b2ac1779ad:/opt/airflow# mypy --namespace-packages airflow/macros
Success: no issues found in 2 source files
```

* Update __init__.py

Part of #19891

Another attempt to clean-up all MyPy errors in Google Provider.

Part of https://github.com/apache/airflow/issues/19891

fixes: #20426 and change in logic introduced (seemingly accidentally) in 2fadf3c3

Part of #19891
There were few more places where I missed adding Sequence
typing - including examples (also converted to tuples) and
also template_ext. Also in a few places iterable was left

Description: allows users to specify they type of secret they are adding when adding extra secrets. Previously we were just defaulting to Opaque.

The new snowflake library version generates a different warning
message as they bumped pyarrow version used. This PR adds the
warning to known warnings.

New Sphinx highlighter does not like spaces in json after ':' and it
fails main builds because of that.

We change the default for `is_delete_operator_pod` to `True`.  For subclasses `GKEStartPodOperator` and `EksPodOperator` we do not _yet_ change the default since we may not want to do a major release in those providers.  Instead we identify when the parameter is not set and emit a deprecation warning to notify users of the impending change.