Improve baseline cost output format to show all clusters

[homatthew]

Summary

The previous format only showed instance info from the first cluster, which was misleading for composite models like Key-Value that have multiple cluster types (Cassandra, EVCache, dgwkv).

• Show all clusters with full details: cluster_type, deployment (zonal/regional), instance, count, annual_cost
• Add attached_drives field per cluster (only when drives exist)
• Remove misleading single instance_name/instance_count/deployment fields

This makes the baseline output self-documenting and easier to review when costs change.

Example: kv_with_cache (before)

{
  "instance_name": "c6id.2xlarge",
  "instance_count": 39,
  "deployment": "zonal",
  ...
}

Example: kv_with_cache (after)

{
  "clusters": [
    {"cluster_type": "evcache", "deployment": "zonal", "instance": "c6id.2xlarge", "count": 39, ...},
    {"cluster_type": "cassandra", "deployment": "zonal", "instance": "c5d.12xlarge", "count": 2, ...},
    {"cluster_type": "dgwkv", "deployment": "regional", "instance": "c7a.2xlarge", "count": 78,
     "attached_drives": ["gp2 : 20GB"], ...}
  ],
  ...
}

Test plan

• tox -e pre-commit passes
• tox -e capture-baseline regenerates with new format
• Regression tests pass

🤖 Generated with Claude Code