Adjust ray operators in google provider. Add ray package as a dependency.

[Crowiant]

• Add ray package as a separate dependency for google provider to avoid downgrading of the setuptools package that happens if ray was installed as google-cloud-aiplatform[ray] ( https://github.com/googleapis/python-aiplatform/blob/main/setup.py#L114 ).
• Remove default values for the fields python_version and ray_version in CreateRayClusterOperator.
• Add documentation regarding the changes.
• Change system test for ray operators.

---

^ Add meaningful description above
Read the Pull Request Guidelines for more information.
In case of fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
In case of a new dependency, check compliance with the ASF 3rd Party License Policy.
In case of backwards incompatible changes please leave a note in a newsfragment file, named {pr_number}.significant.rst or {issue_number}.significant.rst, in airflow-core/newsfragments.

[potiuk]

Is not that contradicitng that ray package is >= 2.42 and that supported ray versions are 2.9.3, 2.33, 2.42 ?

[Crowiant]

Is not that contradicitng that ray package is >= 2.42 and that supported ray versions are 2.9.3, 2.33, 2.42 ?

@potiuk I tested it with ray version 2.47(the latest) in breeze and it works ok. The only warning from vertex ai that it is better to have the same library.
It only impacts the version(field ray_version) that user can use to create a ray cluster in GCP.

[VladaZakharova]

@potiuk
Hi
We were thinking how to make it work in Google provider and in Composer, so this is the solution for now

[eladkal]

I don't like that our docs reference to internal details of other packages - especially when we are referencing main branch where any commit can make this link not accurate. doesn't the library serve a user facing doc that we can reference instead?

[VladaZakharova]

As I understand, this is specifically for users to see the limitations of the package they are trying to use, because without this it will be hard to determine what is wrong if the user will use the operator with unsupported versions of Ray.
For the link to the comment I agree, maybe better to refer to it in another way, more flexible one

[eladkal]

The package has doc it looks like they just don't document which ray version they support. Might worth asking them to do so?
The sentence of

Currently supported versions of ray package in ray cluster are: 2.9.3, 2.33, 2.42.
For more information you can check: https://github.com/googleapis/python-aiplatform/blob/main/setup.py#L101

Should be something like:
To read more about compatible versions of Ray check the Google docs in <Link to upstream doc>

This avoids the need to update our docs for every change upstream library makes.