Must commit changes before reading all nodes on restart

Noticed this when restarting my Holesky node, might be related to `--forceCheckpointSync` with an existing db state as removing the db resolved the issue.

Bit strange though as I haven't observed that before


but what looks suspicious is the debug log, it says that db state is used because it's ahead but with `--forceCheckpointSync` this shouldn't be the case, we always want to use the pulled checkpoint state

```
Mar-29 18:16:21.469[] info: Lodestar network=holesky, version=v1.28.1/cayman/bump-libp2p/44e7e10, commit=44e7e10ba9aa097d686b76fa95c455443e6abd9f
Mar-29 18:16:21.499[] info: Connected to LevelDB database path=/home/devops/holesky/data/beacon2/chain-db
Mar-29 18:16:23.498[] verbose: Found the last archived state slot=3947392, size=433.36 MB
Mar-29 18:16:31.948[] warn: Forced syncing from checkpoint even though db state at slot 3947392 is within weak subjectivity period
Mar-29 18:16:31.949[] warn: Please consider removing --forceCheckpointSync flag unless absolutely necessary
Mar-29 18:16:31.949[] info: Fetching checkpoint state checkpointSyncUrl=https://beaconstate-holesky.chainsafe.io
Mar-29 18:16:46.509[] info: Download completed stateId=finalized, size=433.37 MB, slot=3947328
Mar-29 18:17:10.600[] verbose: Db state is ahead of the provided checkpoint state, using the db state to initialize the beacon chain
```

needs further investigation

Reproduce a Holesky restart with an existing LevelDB state and --forceCheckpointSync, using the logged startup sequence as the starting point. Investigate why the database state is selected after the checkpoint download; done means the behavior is understood and the forced checkpoint path selects the intended state.

Written by the indexing model from the issue text.