fix multi-line buffering issue

[Bradley-Butcher]

Fix subprocess transport handling of buffered JSON output

TL;DR

This PR fixes an issue where the SDK fails to parse output from the Claude CLI when multiple JSON objects are concatenated on a single line due to stdout buffering.

Problem

In certain environments, stdout buffering can cause multiple JSON objects to be delivered as a single line with embedded newlines (\n). When this happens, json.loads() successfully parses the first JSON object but then raises a JSONDecodeError: Extra data error because there's additional valid JSON after the first object.

Example of problematic output:
{"type": "message", "id": "msg1", "content": "First message"}\n{"type": "result", "id": "res1", "status": "completed"}

Solution

The fix modifies the receive_messages() method in SubprocessCLITransport to split each received line on newline characters before attempting JSON parsing. This allows multiple concatenated JSON objects to be properly handled.

Changes

• Modified src/claude_code_sdk/_internal/transport/subprocess_cli.py to split lines on \n before JSON parsing
• Added comprehensive test in tests/test_subprocess_buffering.py to verify the fix

Testing

• Added new test that reproduces the buffering scenario

[ltawfik]

thanks @Bradley-Butcher . Do you mind re-creating the pr with commit signing as it's a requirement to merge the PR, thank you

[Bradley-Butcher]

Should be good @ltawfik

[Mng-dev-ai]

@ltawfik I'm still getting claude_code_sdk._errors.CLIJSONDecodeError: Failed to decode JSON: {"type":"assistant","message":{"id":"msg_01DR48uqZZ3Sru1ttoyG4fD7","type":"message","role":"assistan... after bumping the version to 0.0.13

btw, This PR is working fine for me