refactor(python): unify read/write api for python xlang and native mode

[chaokunyang]

Why?

• Python currently maintains separate serializer APIs (write/read vs xwrite/xread) and separate runtime paths for native vs xlang serialization, which duplicates logic and increases maintenance cost.
• Unifying the Python serializer API and runtime path simplifies custom serializer development and keeps behavior aligned across modes.

What does this PR do?

• Unifies Python serializer APIs to write/read and removes mode-specific xwrite/xread methods from serializer implementations.
• Removes XlangCompatibleSerializer from Python exports and updates custom serializer docs/examples to the unified API.
• Refactors Python runtime serialization/deserialization internals to a single write_ref/write_no_ref and read_ref/read_no_ref path, removing x-prefixed duplicates.
• Unifies dataclass and union serializer implementations/codegen to use the shared read/write path in both xlang and non-xlang modes.
• Makes field_nullable apply regardless of xlang, and updates Python docs accordingly.
• Updates Python tests to validate the unified serializer API and generated code paths.

Related issues

Closes #3347

Does this PR introduce any user-facing change?

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

Benchmark