feat(python/cpp): add streaming serialization support to python and c++

[chaokunyang]

Why?

Add output-stream serialization support so large payloads can be written incrementally instead of materializing full byte arrays first, and align Python/C++ stream APIs for end-to-end stream roundtrips.

What does this PR do?

• Add C++ output-stream serialization APIs: Fory::serialize(OutputStream&, const T&), Fory::serialize(std::ostream&, const T&), and corresponding ThreadSafeFory overloads.
• Unify stream input types in C++ (StreamReader/ForyInputStream to InputStream/StdInputStream) and wire stream-writer state/reset handling through WriteContext and Buffer.
• Introduce flush barrier and incremental flush integration for chunked serializers (notably map and struct paths) to keep chunk headers valid while streaming.
• Add Python stream-write bridge (PyOutputStream), Buffer.wrap_stream + stream writer binding, and Fory.dump(obj, stream) for direct serialization to Python writable streams.
• Expand tests in C++ and Python for stream output roundtrip/parity, partial-write flush behavior, and map chunk streaming paths.

Related issues

Closes #3445
Closes #3444
Closes #3452

Does this PR introduce any user-facing change?

Yes. This adds new public stream serialization APIs in C++ and Python (dump to writable streams).

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

Benchmark