Apply split.dictionary defaults to invalid inputs and null values #1223

Problem

split.dictionary currently uses default only for absent dictionary keys. A non-dictionary input such as the extract.ai status string Timed Out still raises an error. An existing key with a null value overrides its configured default and then becomes a blank cell in ordinary recipe column output.

For example, with default: {Voltage: {}}, {"Voltage": null} should produce an empty dictionary while {"Voltage": {"value": 120, "uom": "VAC"}} should retain the original object.

Agreed workstream

split-dictionary-defaults

Accepted behavior

• An explicitly supplied default, including an empty object, enables fallback for invalid or non-dictionary input cells. Treat those cells as empty dictionaries while retaining valid values from other input columns.
• Preserve errors for invalid inputs when no default is supplied, and do not hide invalid default configuration.
• Apply per-key defaults to missing keys and explicit null values after merging all input dictionaries. Later valid input dictionaries retain their existing precedence.
• Preserve non-null values, including zero, false, empty strings, lists, and dictionaries. Keys remain case-sensitive, defaults use original names before output renaming, and nested values are not filled recursively.
• Apply the same rules to columns and to_lists. Copy mutable fallback values independently for each output cell.

Acceptance checks

• Reproduce the timeout-string case and a convert.from_json followed by split.dictionary recipe containing valid Voltage objects, a null Voltage, and a missing Voltage.
• Cover strict behavior without a default, empty and invalid defaults, multi-input precedence, preserved non-null values, renaming/filtering, and mutable-default isolation.
• Update the generated-schema source descriptions and validate the recipe schema.
• Run the credential-free local suite; report any live XL/service validation separately.