[Improvement]feat(operator): allow update replica of rss shuffle server when upgrading

### Code of Conduct

- [X] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct)

### Search before asking

- [X] I have searched in the [issues](https://github.com/apache/incubator-uniffle/issues?q=is%3Aissue) and found no similar issues.

### What would you like to be improved?

Currently, rss operator performances the upgrading of StateFullSet of shuffle server incrementally and only updates the rss object from upgrading to running when the sts is ready.

However for increasing replica of shuffle server, there might be cases that the replica is too large that K8S cluster doesn't have enough resources to scheduling all the shuffle server, which means sts can never be ready, which makes the rss object hanging in the `upgrading` status forever.

From the webhook side of rss operator, it forbids any update to an `upgrading` rss object.

Considering these two facts together, rss obejct could be ended up in `upgrading` state and there's currently no way to change that from the rss object side.

### How should we improve?

I'd like to propose that allowing modify replica of shuffle server even when the rss object is upgrading. Therefore, admins of rss cluster could decrease the replica and makes the rss object entering the running state.

### Are you willing to submit PR?

- [ ] Yes I am willing to submit a PR!