Handle malformed API response values in model parsing

[IFAKA]

Problem

The AnalysisResult.from_dict() method uses direct float() and int() conversions without error handling:

confidence=float(classification.get("confidence", 0)),
word_count=int(analysis.get("word_count", 0)),

If the API returns malformed data (non-numeric strings, None when key exists, etc.), the SDK crashes with ValueError:

>>> AnalysisResult.from_dict({"classification": {"confidence": "invalid"}})
ValueError: could not convert string to float: 'invalid'

Fix

Add safe conversion helpers that return defaults on failure:

def _safe_float(value: Any, default: float = 0.0) -> float:
    if value is None:
        return default
    try:
        return float(value)
    except (ValueError, TypeError):
        return default

Test

Test 1: Valid data
  ✓ Parsed correctly: confidence=0.85, word_count=10

Test 2: Non-numeric strings
  ✓ Handled gracefully: confidence=0.0, outlook=0.0

Test 3: None values
  ✓ Handled None: confidence=0.0

Test 4: Missing keys
  ✓ Handled missing keys: confidence=0.0

Test 5: Empty dict
  ✓ Handled empty dict: label=neutral

==================================================
All tests passed!