[FEATURE] Support stage recompute for Tez clients

### 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.

### Describe the feature

For tez without remote shuffle server, when task recieve too many fetch exception from downstream task, this task will be recomputed.
With remote shuffle server, if some shuffle server is down, will loss data. Because data have been merged from many upstream task, so we must recompute upstream vertex.

This issue is the problem 2.d which is described in #1011

### Motivation

_No response_

### Describe the solution

_No response_

### Additional context

_No response_

### Are you willing to submit PR?

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

Start by reading issue #1011, especially problem 2.d, and trace how Tez clients handle downstream fetch exceptions with and without the remote shuffle server. Define the required stage-recompute behavior when a shuffle server is unavailable, then identify the affected entry points and tests. Done means upstream vertices can be recomputed without losing merged shuffle data.

Written by the indexing model from the issue text.