Desktop DB graph download failures are only logged and leave misleading progress UI

Nobody has claimed this yet.

### Search first

- [x] I searched open and closed issues and found no Desktop report focused on error surfacing for this flow.

### What happened?

General DB graph snapshot-download failures are written to Electron or graph-local worker logs but are not presented as actionable UI errors.

Two reproduced examples:

1. Worker runtime startup failed. The last visible message remained `Preparing graph snapshot download` while the Electron log contained `db-worker-node failed to start`.
2. The worker started, but the initial pull failed with a TLS error. The UI briefly showed `Downloading...`, then stopped and allowed an empty graph to be opened. The graph-local worker log contained:

```text
:db-sync/download-graph-by-id-failed
:stage :fetch-pull
TypeError: fetch failed
```

The operation can therefore look hung or completed even though snapshot transfer never began.

### Reproduce

1. Initiate a remote DB graph download.
2. Force either a worker startup failure or a network failure at `:fetch-pull`.
3. Observe the progress UI and compare it with Electron main and graph-local worker logs.

### Expected behavior

- Clear the progress state on every failure path.
- Show the failed stage and a sanitized underlying error.
- Do not switch into or present a failed empty graph as usable.
- Offer Retry and Remove failed local graph actions where appropriate.

### Platform

- Logseq Desktop 2.0.1
- macOS arm64
- DB Sync and db-worker-node

Trace the remote DB graph download flow across the Electron main process and graph-local worker, starting at the db-worker-node startup failure and :db-sync/download-graph-by-id-failed with :stage :fetch-pull. Reproduce startup and TLS failures, then verify that progress is cleared, the failed stage and sanitized error are shown, failed empty graphs are not usable, and Retry or Remove actions appear where appropriate.

Written by the indexing model from the issue text.