Update TokenAwarePolicy.make_query_plan to schedule to replicas first

[dkropachev]

End result should be like this:

1. Replicas from the same RACK (if rack is specified)
2. Replicas from the same DC, but Remote RACK (if rack is specified)
3. Replicas from the same DC (if rack is not specified)
4. Replicas from the remote DC
5. Non-replicas in the same order

Pre-review checklist

• I have split my patch into logically separate commits.
• All commit messages clearly explain what they change and why.
• I added relevant tests for new features and bug fixes.
• All commits compile, pass static checks and pass test.
• PR description sums up the changes and reasons why they should be introduced.
• I have provided docstrings for the public items that I want to introduce.
• I have adjusted the documentation in ./docs/source/.
• I added appropriate Fixes: annotations to PR description.

[mykaul]

Shouldn't we add tests to ensure this works well?

[dkropachev]

Shouldn't we add tests to ensure this works well?

Of course.

[sylwiaszunejko]

Replicas from the same RACK (if rack is specified)
Replicas from the same DC, but Remote RACK (if rack is specified)
Replicas from the same DC (if rack is not specified)
Replicas from the remote DC
Non-replicas in the same order

Are we sure that this is the right order? I vaguely remember a discussion during summit whether we should query local (dc/rack) nodes that are non-replicas or remote replicas first, and the answer was not obvious. @Lorak-mmk @avikivity

[dkropachev]

@gusev-p we would appreciate your opinion on this matter too.

[dkropachev]

@gusev-p , @avikivity , @Lorak-mmk , please take a look, we need to release python-driver ASAP

[Lorak-mmk]

I don't see a point in waiting for discussion about the correct order. The PR improves the order undoubtedly, and it matches e.g. Rust Driver order (and probably some other drivers too).
Changing this order is imo a separate task and we can discuss it in https://github.com/scylladb/scylla-drivers/issues/36