feat(core): compress ValidMeasurementResults when serializing

[AlessandroPomponio]

Summary

This PR implements a compressed serialization format for ValidMeasurementResult to eliminate redundant ExperimentReference data. Instead of repeating the same ExperimentReference in every measurement's property, it's now stored once at the top level, significantly reducing serialized size. The implementation maintains full backward compatibility with the old format.

Resolves #709

Files Changed

📄 orchestrator/schema/result.py

Added custom serialization and deserialization methods to ValidMeasurementResult class:

• serialize_to_custom_format(): Custom serializer that extracts the common ExperimentReference (validated to be identical across all measurements) and stores it at the top level, removing it from individual ObservedPropertyValue entries. Properly handles all Pydantic serialization settings like exclude_none, exclude_defaults, etc.
• deserialize_from_custom_format(): Custom validator that handles both old (redundant) and new (compressed) formats, ensuring backward compatibility. Detects the format and reconstructs measurements appropriately.
• Removed outdated docstring comment about accessing experiment reference through measurements.

📄 tests/schema/test_measurement_result.py

Added comprehensive test coverage for the new serialization format:

• test_compressed_serialization_format(): Verifies the new format structure with experimentReference at top level
• test_old_format_deserialization(): Ensures backward compatibility by testing deserialization of old format data
• test_new_format_deserialization(): Validates correct deserialization of the new compressed format
• test_serialization_deserialization_roundtrip(): Tests full roundtrip to ensure data integrity
• test_compression_achieved(): Quantifies compression benefits by comparing serialized sizes
• test_json_serialization_roundtrip(): Verifies JSON string serialization/deserialization works correctly with the new format

[AlessandroPomponio]

I confirmed the minimization is happening by running a new operation using https://ibm.github.io/ado/examples/no-priors-characterization/

I was then able to do ado show entities with both the latest operation and one that was run before this change

[michael-johnston]

Nice - looks clean.