feat: introduce initial snapshot chunking

[icehaunter]

This PR changes the way we write snapshots to disk, the storage API to request them, and some logic around offsets.

1. The Storage.get_snapshot/1 method is gone. It's now unified under Storage.get_log_stream/3, with some caveats.
2. The offsets within the snapshot rows are not updated, so now latest item on the sent log may have mismatching offset field to the electric-offset header of the same response. All our clients respect the header, and that's the only correct behaviour going forward. There is a decision to remove offset from our items entirely, but that's a separate PR.
3. While we're making the snapshot, we cannot know how many chunks are going to be there until we make them. Thus we allow clients to request any chunks up to LogOffset.new(0, :infinity), and consider all offsets at txn 0 to be virtual pointers.
4. The get_log_stream/3 for the snapshot section ignores the contract a little bit - you cannot request the stream to be up to an arbitrary point, it's always up to the chunk boundary, even when specified otherwise. Since we've moved to "pure" chunking behaviour, it should be enshrined in the Storage interface so that reading functions can act optimally and honestly. I'm hoping to address that in a separate PR
5. If the instance is started with the old snapshot already created, we update the naming and metadata to avoid recreating shapes if we can avoid it.

[netlify]

✅ Deploy Preview for electric-next ready!

Name	Link
🔨 Latest commit	c531cfc
🔍 Latest deploy log	https://app.netlify.com/sites/electric-next/deploys/67568b3107c1db00082d5e10
😎 Deploy Preview	https://deploy-preview-2119--electric-next.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[icehaunter]

@msfstef can you please take another look at .skip-ed tests in the TS client that deal with headers and confirm that this me disabling them is correct? The general idea is "well we're serving only full chunks when serving snapshot, and up to file boundary, so switching to 'live' log is a separate HTTP request, hence e-tags should no longer mutate because response no longer mutates"

[msfstef]

Very nice work! Love the ripping out of snapshot logic in the serve plug!

Left some comments and questions - as for the TS tests checking etags, we can modify them to be looking at the first "incomplete" actual log chunk rather than looking at -1 (which used to collate the transaction log)