feat: support reset dead partition

## Feature Request

**Is your feature request related to a problem? Please describe:**
When data disk damage causes all three replicas to be unavailable, most of Pegasus's functions—such as backup, bulk load, and duplication—will not be usable. Therefore, a fallback strategy is needed to address this situation.

**Describe the feature you'd like:**

By resetting the dead paritions to abandon some data, you can trick the meta server into allowing functions like backup to proceed normally.

**Mainly two parts:**
1. Reset the partition configuration on the meta server and zookeeper.
2. Reserve the dead partition data on the target replica server.

No files, tests, or entry points are identified. Start by tracing how the meta server and ZooKeeper store partition configuration, then examine how target replica servers retain dead-partition data. Done means both configuration reset and dead-data reservation work together so backup, bulk load, and duplication can proceed.

Written by the indexing model from the issue text.