Sigma converter silently drops key columns when a relationship's from_columns/to_columns lengths differ

Start in converters/sigma at OssieToSigmaConverter._build_relationship and reproduce the unequal from_columns/to_columns case from the issue. Check how relationship keys and issues are represented, then ensure the extra key column is not silently lost and that the caller can see the mismatch; rerun the provided conversion example to verify the result.

Summary

OSIToSigmaConverter._build_relationship pairs a relationship's from_columns and to_columns with a bare zip(). The OSI schema only requires each array to have at least one entry independently ($defs/Relationship in core-spec/ossie-schema.json); it does not require them to be the same length. So a compound-key relationship with unequal-length arrays is legal input, and zip() silently stops at the shorter array, dropping the extra key column(s) from the exported Sigma spec with no warning.

Steps to reproduce

from ossie import OssieDataset, OssieDocument, OssieRelationship, OssieSemanticModel
from ossie_sigma.ossie_to_sigma import OssieToSigmaConverter

document = OssieDocument(
    semantic_model=[
        OssieSemanticModel(
            name="m",
            datasets=[
                OssieDataset(name="orders", source="db.public.orders"),
                OssieDataset(name="regions", source="db.public.regions"),
            ],
            relationships=[
                OssieRelationship(
                    name="OrderRegion", **{"from": "orders"}, to="regions",
                    from_columns=["region_id", "sub_id"], to_columns=["region_id"],
                ),
            ],
        )
    ]
)
result = OssieToSigmaConverter().convert(document)
print(result.output["pages"][0]["elements"][0]["relationships"][0]["keys"])
print(result.issues)

keys comes out with a single region_id pair; sub_id is gone and issues is empty, so nothing about the loss is visible to the caller.

Environment

apache/ossie main @ fc6c9df (2026-09-15), converters/sigma.