Add special handling for snapshot root slot in get_confirmed_block

[CriesofCarrots]

Problem

Testnet panics when downstream user calls getConfirmedBlock rpc on all slots. (And when all the rpc threads have panicked due to retries, the rpc node becomes unresponsive to further queries.)

In particular, the panic occurs on getConfirmedBlock on the first blocktree root (aside from slot 0). Blocktree::get_confirmed_block assumes a root always has a valid parent slot, but when a node is booted from a snapshot (as the rpc node is on testnet boot), the snapshot root doesn't contain any entries or have a valid parent root in blocktree.

Summary of Changes

• Add handling for snapshot root slot (no parent root, no entries; returns BlocktreeError::SlotNotRooted), and next slot (parent slot has no entries, and therefore no blockhash; returns previous_blockhash: Hash::default())

Technically, this means the results of getConfirmedBlock for snapshot root and snapshot root+1 are slightly incorrect. But we expect our main downstream use case to run a node without snapshots, so this should not be an issue.
If it becomes an issue, the more general solution would be to preserve the entries of snapshot root in each snapshot.

This should fix testnet.solana.com, when deployed there.

Fixes #7418

[mvines]

r+ but no test? 😬

[codecov]

Codecov Report

Merging #7430 into master will increase coverage by 6.4%.
The diff coverage is 97.8%.

@@           Coverage Diff            @@
##           master   #7430     +/-   ##
========================================
+ Coverage    74.2%   80.6%   +6.4%     
========================================
  Files         245     245             
  Lines       52559   48466   -4093     
========================================
+ Hits        39003   39075     +72     
+ Misses      13556    9391   -4165

Pull request has been modified.

[CriesofCarrots]

r+ but no test? 😬

Had hoped to beat you there!