[release/6.x] Respond with available snapshot rather than Bad Request error

[eddyashton]

Partial cherry-pick from #7373 to resolve #7409.

#7373 changed behaviour on both the client and server side to support the new flow. The problem is that the client may now send requests that cause an old (6.x) server to respond with a 400 Bad Request, and it seems downstream of that we occasionally fail the LTS test.

This cherry-picks the server-side change, so it will accept these request and respond with the sub-range.

We could revert this behaviour change on main, and then try another multi-step migration. I think it'd still need a change on 6.x to match up. We could also change the behaviour on main/7.x to fallback when we receive a 400 and interop nicely with a known-old node. I think that's a huge amount of extra code and complication, and not worth it. So this is my proposed fix.

[achamayou]

@eddyashton I think this requires cutting a 6.0.16 to fix LTS, and updating the changelog to make it clear that updates from <6.0.16 to 7.0.0 are unsafe, out of abundance of caution.

[eddyashton]

I think this requires cutting a 6.0.16 to fix LTS

Agreed

updating the changelog to make it clear that updates from <6.0.16 to 7.0.0 are unsafe

I don't think we need this - standard guidance is that 7.0.0 (which doesn't exist yet) requires 6.latest, and the breakage when these interop is non-fatal.