Fix eternally paused state sync and chunk ordering issues

[paberr]

What's in this pull request?

In combination with validity sync issues, we saw state sync getting stuck. This could happen if we synced all chunks (end_key = None) and subsequently paused state sync but then failed to apply the last chunk.

This PR addresses the issue in two ways:

• Store the block number at which the last chunk should be applied. If we pass it and are not complete, unpause the state sync.
• If a chunk was still buffered from before a reset, a new duplicate chunk with the same start key would be discarded – resulting in an unordered list of chunks that eventually will fail to apply. We thus replace the duplicate check by a check that the buffer is ordered. It it isn't, we discard old entries in favour of new ones.

Pull request checklist

• [x] All tests pass. The project builds and runs.
• [x] I have resolved any merge conflicts.
• [x] I have resolved all clippy and rustfmt warnings.

[viquezclaudio]

I tested this fixes the validity sync stuck issue