test(local-cluster): liveness after intertwined skip-fallback and notar-fallback

[ksn6]

Problem

Test to validate Alpenglow's ability to maintain liveness when nodes issue both NotarizeFallback and SkipFallback votes in an intertwined manner.

This test simulates a consensus scenario with four nodes having specific stake distributions:

• Node A: 40% + epsilon stake
• Node B: 40% - epsilon stake
• Node C: 20% - epsilon stake
• Node D: epsilon stake (minimal, acts as perpetual leader)

The test proceeds through two main stages:

Stage 1: Stable Network Operation

All nodes are voting normally for leader D's proposals, with notarization votes going through successfully and the network maintaining consensus.

Stage 2: Network Partition and Fallback Scenario

At slot 50, Node A's turbine is disabled, creating a network partition. This triggers the following sequence:

1. Node D (leader) proposes a block b1
2. Nodes B, C, and D can communicate and vote to notarize b1
3. Node A is partitioned and cannot receive b1, so it issues a skip vote
4. The vote distribution creates a complex fallback scenario:
   • Nodes B, C, D: Issue notarize votes initially, then skip fallback votes
   • Node A: Issues skip vote initially, then notarize fallback vote
5. This creates the specific vote pattern:
   • B, C, D: notarize + skip_fallback
   • A: skip + notarize_fallback

The test validates that:

• The network can handle intertwined fallback scenarios
• Consensus is maintained despite complex vote patterns
• The network continues to make progress and create new roots after the partition is resolved
• At least 10 new roots are created post-experiment to ensure sustained liveness

[bw-solana]

This is a great test with awesome documentation. Thank you!

Left a handful of nits and a couple suggestions

[bw-solana]

LGTM