Trie ChangSets: Fix pipeline checkpoints

[mediocregopher]

Describe the feature

There are two bugs regarding the pipeline checkpoint for the new MerkleChangeSets stage:

• Inside of the OverlayStateProviderFactory we are checking the checkpoint range within the StageCheckpoint, but in reality only the outer block_number field gets updated by the engine when processing blocks.

• When running on a dev box I noted that the pipeline sync stage is running with a massive range:

2025-10-08T09:49:23.823777Z  INFO Status connected_peers=130 stage=MerkleChangeSets checkpoint=23503319 target=23526851

This is because the pipeline was restarted after the node was down for a time. 23503319 was indeed the finalized block at the time the node was shut down, but we should limit the sync stage so as to not end up in this situation.

Additional context

No response