API uses naive [0] index for multiple P580/P582 qualifiers in timeframe comparison

Problem

When comparing position timeframes during import/enrichment, the API extracts only the first qualifier value ([0]) for P580 (start date) and P582 (end date). This is problematic because ~389 positions have multiple P580 values and similar numbers have multiple P582 values.

Affected Code

poliloom/models/politician.py:934-960:

@staticmethod
def _extract_timeframe_from_qualifiers(qualifiers_json: dict | None):
    if "P580" in qualifiers_json:
        start_data = qualifiers_json["P580"][0]["datavalue"]["value"]  # <-- [0] only
        start_date = WikidataDate.from_wikidata_time(...)
    if "P582" in qualifiers_json:
        end_data = qualifiers_json["P582"][0]["datavalue"]["value"]  # <-- [0] only

This method is used by:

• Property._compare_to() - comparing properties during import
• Property.should_store() - deciding whether to store new properties
• Property.soft_delete_matching_extracted() - deduplication during import

Additionally, Property.format_timeframe() (lines 893-931) also uses [0] for display purposes.

Impact

Incorrect timeframe extraction can cause:

1. False matches - positions incorrectly identified as duplicates
2. Missed matches - overlapping positions not detected
3. Inconsistent display - arbitrary date shown instead of most representative

GUI Solution (for reference)

The GUI solved this in poliloom-gui/src/lib/wikidata/qualifierParser.ts with selectBestDate():

1. Parse all date values from the qualifier array
2. Sort by precision (most precise first: day > month > year)
3. For equal precision: earliest for P580, latest for P582
4. Return the best candidate

Proposed Fix

Port the GUI's selectBestDate logic to Python's _extract_timeframe_from_qualifiers() method:

@staticmethod
def _select_best_date(snaks: list, prefer_latest: bool = False) -> WikidataDate | None:
    """Select best date from multiple qualifier values.
    
    Strategy: prefer most precise date, then earliest for start dates
    or latest for end dates.
    """
    parsed_dates = []
    for snak in snaks:
        if "datavalue" in snak and "value" in snak["datavalue"]:
            value = snak["datavalue"]["value"]
            date = WikidataDate.from_wikidata_time(value["time"], value["precision"])
            if date:
                parsed_dates.append(date)
    
    if not parsed_dates:
        return None
    if len(parsed_dates) == 1:
        return parsed_dates[0]
    
    # Sort by precision (descending), then by date
    def sort_key(d):
        python_date = d.to_python_date()
        date_value = python_date.toordinal() if python_date else 0
        if prefer_latest:
            date_value = -date_value
        return (-d.precision, date_value)
    
    parsed_dates.sort(key=sort_key)
    return parsed_dates[0]

Related

• Fixes part of the issue identified in #73
• GUI fix: poliloom-gui/src/lib/wikidata/qualifierParser.ts

Start with poliloom/models/politician.py:893-960, especially _extract_timeframe_from_qualifiers(), and compare the existing selection behavior with poliloom-gui/src/lib/wikidata/qualifierParser.ts and selectBestDate(). Done means timeframe comparison, storage, deduplication, and display consistently choose dates by precision and the stated start/end tie-breaking rules.