Is it mandatory to configure table joins exclusively through semantic models when defining metrics in MetricFlow?

[lach1n3]

In our use case, we need more flexibility in configuring table joins beyond what's defined in semantic model YAML files. We want to:

• Maintain existing semantic model configurations (measures, dimensions, etc.)
• Introduce separate configuration artifacts to customize join logic (join types, multiple hop relationships, etc.)
• Combine both configurations to generate unified metrics/measures for queries

Does metricflow support the above requirements?