fix: respect default_factory for abstract generic fields in from_dict

[sanjibani]

Problem

When a dataclass field is typed as an abstract generic collection (e.g. MutableMapping[str, str]) and given a default_factory, calling from_dict({}) raises TypeError: MutableMapping() takes no arguments.

Fixes #505.

Root cause

In _decode_dataclass, fields missing from the input dict are correctly populated via default_factory(). However, the subsequent decoding loop still calls _decode_generic on that already-correct value, which attempts to instantiate the abstract MutableMapping type directly.

Fix

Track which fields were populated by default_factory (as opposed to arriving from the input data). In the decode loop, skip _decode_generic for those fields and pass the factory-produced value through as-is.

Test

Added a regression test in tests/test_collections.py covering both the empty-dict case and a full round-trip:

```python
@DataClass
class MyClass(DataClassJsonMixin):
field1: MutableMapping[str, str] = field(default_factory=dict)

MyClass.from_dict({}) # previously raised TypeError
MyClass.from_dict(MyClass(field1={"key": "value"}).to_dict()) # round-trip
```

Full test suite: 323 passed, 4 skipped, 0 failures.

🤖 Generated with Claude Code