Support calculated measures (derived metrics) in YAML

[boringdata]

Problem

BSL supports calc_measures in the Python API (SemanticModel(calc_measures={...})) but not in YAML definitions. This means derived metrics like fraud_rate = fraud_volume / transaction_volume must inline the full expression:

fraud_rate:
  expr: (_.has_fraudulent_dispute.cast("int64") * _.eur_amount).sum() / _.eur_amount.sum()

Instead of the more maintainable:

calculated_measures:
  fraud_rate:
    expr: fraud_volume / transaction_volume
    description: "Fraud rate as ratio of fraudulent volume to total volume"

Proposal

Add a calculated_measures section to the YAML schema that allows referencing other measures by name. This would:

1. Parse calculated_measures in from_yaml() / from_config()
2. Resolve measure references at evaluation time
3. Pass them through to SemanticModel(calc_measures=...)

Why

• DRY: Avoid duplicating aggregation logic across measures
• Readability: fraud_volume / transaction_volume is clearer than the inlined version
• Consistency: The Python API already supports this via calc_measures param
• LLM agents: Agents reading semantic_model.yml can better understand metric relationships when they see explicit references