Dataset deduplication with live progress feedback in Channel view

### StatGPT Admin Frontend version

latest

### What is the problem this feature will solve?

Currently, a user can start a deduplication process for a channel, but has no visibility into whether the process is running, how far along it is, or how it ended. Deduplication is an asynchronous backend job that can take several minutes, so a single fire-and-forget request gives the user no confidence that anything is happening and provides no summary of what was changed.

### What is the feature you are proposing to solve the problem?

Show a live, status-specific notification as the deduplication job moves through its lifecycle.
Polls the backend for the latest job status until it reaches a terminal state.
On completion, shows a summary of what changed (records remapped/deleted); on failure, shows the backend's reason; if the job runs unusually long, informs the user it's continuing in the background.

### What alternatives have you considered?

_No response_

Start by locating the Channel view's existing deduplication request and identify how the backend exposes job status and terminal results. Trace the lifecycle from starting the job through polling, completion, failure, and long-running states; done means each state presents the specified notification and completion includes the records remapped or deleted.

Written by the indexing model from the issue text.