Why?

This PR adds field-level metadata configuration support for Python, completing the cross-language field metadata feature across all Fory language implementations (Java, Rust, Go, C++, and now Python).

Field metadata allows users to:

• Use numeric tag IDs instead of field names for more compact serialization
• Control nullable flags, reference tracking, and field ignoring per field
• Enable schema evolution with stable field identifiers

What does this PR do?

1. New pyfory.field() API

Introduces a new pyfory.field() function for fine-grained control over serialization behavior per field:

from dataclasses import dataclass
from typing import Optional, List
import pyfory
from pyfory import Fory, int32

@dataclass
class User:
    id: int32 = pyfory.field(0)                          # Tag ID 0
    name: str = pyfory.field(1)                          # Tag ID 1
    email: Optional[str] = pyfory.field(2, nullable=True) # Tag ID 2, nullable
    friends: List["User"] = pyfory.field(3, ref=True, default_factory=list)  # Tag ID 3, ref tracking
    _cache: dict = pyfory.field(-1, ignore=True, default_factory=dict)  # Field name encoding, ignored

2. TAG_ID Encoding Support

Implements TAG_ID encoding in the xlang serialization protocol:

• id >= 0: Uses numeric tag ID (2-bit encoding = 0b11)
• id = -1: Uses field name with meta string encoding
• More compact than field name encoding
• Stable across field renames

3. Field Header Format

Field header (8 bits):
- 2 bits: encoding type (0b00-10 = field name, 0b11 = TAG_ID)
- 4 bits: size/tag_id (0-14 inline, 15 = overflow)
- 1 bit: nullable flag
- 1 bit: ref tracking flag

4. Schema Evolution Support

When deserializing data with a different schema than the registered class:

• TypeDef meta contains the sender's field information
• TAG_ID is resolved back to actual field names using the receiver's class metadata
• Enables forward/backward compatibility

5. Files Changed

New files:

• python/pyfory/field.py: pyfory.field() function and ForyFieldMeta dataclass

Modified files:

• python/pyfory/struct.py: Updated DataClassSerializer to support field metadata
• python/pyfory/meta/typedef.py: Added TAG_ID to field name resolution
• python/pyfory/meta/typedef_encoder.py: TAG_ID encoding support
• python/pyfory/meta/typedef_decoder.py: TAG_ID decoding support
• python/pyfory/__init__.py: Export field function

Test files:

• python/pyfory/tests/test_field_meta.py: Comprehensive tests for field metadata

Related issues

#3002

Does this PR introduce any user-facing change?

Yes, introduces the new pyfory.field() API for field-level metadata configuration.

• Does this PR introduce any public API change?
• Does this PR introduce any binary protocol compatibility change?

The binary protocol changes (TAG_ID encoding) are already part of the xlang specification and implemented in other languages.

Benchmark

No performance regression expected. TAG_ID encoding is more compact than field name encoding.