fix: SegmentHeader block range is now an Option<_>

[joshieDo]

When a SnapshotProviderRW creates a new file, it commits the configuration before it can be used. And block_range was being initialized as block.start..=block.start.

However, there is the case that if the node is interrupted before the first commit of a new file, then on restart, the node would return the wrong highest block of the segment. It would return block.start, when in fact it's just an empty file.

This would have consequences when calling get_snapshot_target, where it would assume that we need block.start + 1 -> tip data.

Changing SegmentHeader block_range to an option should fix this.

This will cause existing nodes on feat/static-files to be incompatible with this change, since the configuration file format changed
To make this less painful, I'll create a small cli tool that can generate configuration files (conf) with the same ranges, and so they can replace the existing ones.

--

• Additional fix where we make a consistency heal but weren't committing with fsync to disk

[onbjerg]

you mentioned that we can apply this fix w/o resyncing, can you document the procedure for that in the PR?

[shekhirin]

LGTM, one metrics nit