Preserve Matrix type metadata for empty values

[kevinherron]

Summary

Allow zero-length structured and enumerated Matrix values to retain their declared type metadata across decode, encode, and transform paths.

• Add explicit Matrix DataTypeId support so empty structured/enumerated matrices do not infer concrete type metadata from element 0.
• Preserve declared DataTypeIds when constructing structured matrices in binary, JSON, XML, and dynamic codec paths.
• Add typed Matrix transform overloads for empty matrix conversions where the transformed element type cannot be inferred from data.
• Add regression coverage for zero-dimension structured matrix decoding and empty Matrix metadata behavior.

Key Changes

• Matrix metadata: Matrix can now carry an explicit ExpandedNodeId for the concrete element DataType. Empty structured and enumerated arrays no longer attempt to read element 0 during DataTypeId derivation.

• Decode paths: Structured matrix decoders pass the declared DataTypeId into Matrix construction so empty values keep the schema type known by the caller.

• Transform paths: Matrix transformations used by XML encoding and dynamic field codecs can now provide explicit output metadata. This keeps empty transformed matrices valid without changing non-empty inference behavior.

• JSON struct codec: Enum and structured matrices created from JSON field definitions preserve the declared DataTypeId before being passed to downstream encoders.

References

• OPC 10000-6, 5.2.5: Matrix values with zero dimensions may encode no element values.
• Related hardening: Harden binary Matrix dimension decoding #1784