Support `RollingUpdate` deployment strategy

[whynowy]

Summary

Today, when there's any vertex pod recreation needed, we do a Recreate strategy, which means, we clean up all the old pods, and recreate everybody. This will cause some downtime for data processing because pod creation and initialization take time. To be able to achieve more smooth deployment, we should also support RollingUpdate strategy like Deployment does, and the default strategy should be RollingUpdate.

---

Message from the maintainers:

If you wish to see this enhancement implemented please add a 👍 reaction to this issue! We often sort issues this way to know what to prioritize.