[forge] Add mainnet-like P90 latency test as land blocking default

[danielxiangzl]

Summary

• Replaces the land_blocking forge suite with a latency-focused test that uses a mainnet-representative validator distribution (~70% EU, ~20% NA, ~10% Asia)
• The previous default used an even 4-way split (25% each region), which over-weights Asia and under-weights EU relative to actual mainnet (~70% EU, ~25% NA, ~2% Asia)
• The new distribution causes EU proposers to dominate rounds as they do on mainnet, exercising the actual P90 latency bottlenecks (distant proposers racing with EU batch arrival)
• Asia is intentionally over-represented (2% → 10%) so inter-continental tail latency is exercised even with only 20 validators

Test plan

• Trigger Actions → Adhoc Forge with default FORGE_TEST_SUITE=land_blocking to validate the test runs and latency thresholds are reasonable (P50 ≤1.5s, P90 ≤2.5s, P99 ≤4.0s at 3500 TPS)

🤖 Generated with Claude Code

[github-actions]

✅ Forge suite realistic_env_max_load success on 825c202056fc22720d13fc2cfd68722f16de2a92

performance benchmark : committed: 3499.99 txn/s, latency: 281.60 ms, (p50: 200 ms, p70: 300, p90: 300 ms, p99: 800 ms), latency samples: 28540
Latency breakdown for phase 0: ["MempoolToBlockCreation: max: 0.140, avg: 0.104", "ConsensusProposalToOrdered: max: 0.090, avg: 0.088", "ConsensusOrderedToCommit: max: 0.017, avg: 0.015", "ConsensusProposalToCommit: max: 0.107, avg: 0.103"]
Max non-epoch-change gap was: 1 rounds at version 5415 (avg 0.00) [limit 10], 1.12s no progress at version 788501 (avg 0.05s) [limit 30].
Max epoch-change gap was: 0 rounds at version 0 (avg 0.00) [limit 10], 0.00s no progress at version 0 (avg 0.00s) [limit 30].
Test Ok

• Grafana dashboard
• Humio Logs
• Axiom Logs
• Validator CPU Profile
• Fullnode CPU Profile
• Test runner output
• Test run is land-blocking