Refactor processing functions to use RawProcessingResult for less verbose error/warning handling

[Copilot]

resolves #3857

processing_functions.py had ~50 instances of verbose ProcessingAnnotation.from_fields(input_fields, [output_field], AnnotationSourceType.METADATA, message=...) boilerplate. Processing functions don't need to know about annotation structure—only the message matters. The caller (call_function) already has all the context to build the full annotation.

Changes

• datatypes.py: Added RawProcessingResult(datum, warnings: list[str], errors: list[str]) and helpers processing_error(msg) / processing_warning(msg, datum)
• processing_functions.py:
  • call_function now accepts both ProcessingResult and RawProcessingResult, converting the latter by attaching field/type metadata
  • All 13 processing functions converted to return RawProcessingResult
  • Helper functions (check_latin_characters, missing_taxonomy_service_error, taxonomy_network_error) simplified accordingly
  • Removed dead invalid_value_annotation function
  • 2208 → 1719 lines (−22%)
• Tests: Updated assertions from .errors[0].message → .errors[0] (now plain strings)

Before / After

# Before (6+ lines per error)
return ProcessingResult(
    datum=None, warnings=[],
    errors=[ProcessingAnnotation.from_fields(
        input_fields, [output_field], AnnotationSourceType.METADATA,
        message="Something went wrong.",
    )],
)

# After (1 line)
return processing_error("Something went wrong.")

Screenshot

N/A — backend refactoring only, no UI changes.

PR Checklist

• All necessary documentation has been adapted.
• The implemented feature is covered by appropriate, automated tests.
• Any manual testing that has been done is documented (i.e. what exactly was tested?)
  • All 58 preprocessing unit tests pass (excluding test_nextclade_preprocessing.py which requires external nextclade3 binary).