Reject duplicate column names when building record values from `IDataReader`

[Seddryck]

Problem

BuildRecordValue passes each IDataReader.GetName(i) value to RecordValue.Set.

When the reader exposes duplicate column names, the later value replaces the earlier value in the backing dictionary. This silently loses data.

Expected behavior

Validate column names before calling RecordValue.Set and reject duplicate names explicitly.

A duplicate column name should throw a FormatException rather than overwrite an existing field.

Acceptance criteria

• Duplicate names returned by IDataReader.GetName(i) are detected before values are inserted into the record.
• A duplicate name raises FormatException.
• Existing behavior for unique column names remains unchanged.
• Add a reader-normalization regression test using duplicate GetName values.

[Seddryck]

Reviewed as part of 2.0