update Python implementation to match JS

[haneric00]

Modified how we grab initial attributes in ChatModelStart and LLMStart

• separated the 1 shared function for this into two separate functions, one for each parameter we use (serialized and metadata)
• added some extra logic to getNameFromCallback to reduce duplicate code (extra if check for name value)
• modified init.py (bug fix, accidentally deleted the code to remove handlers upon unwrapping)
• added stop sequence to span attributes

IMPORTANT CHANGES:
chat and llm start spans are now named after the API we use ('chat ChatBedrock' or 'chat BedrockLLM' for example) rather than the model used.

• technically against the semantic convention, probably going to change it later but at the current moment this is done to match the JS implementation

[Copilot]

Pull Request Overview

This PR updates the Python implementation to match the JavaScript implementation for LangChain v2 instrumentation. The changes modify how initial attributes are handled in ChatModelStart and LLMStart events and rename spans to use API names instead of model names.

• Refactored attribute extraction into separate functions for serialized and metadata parameters
• Modified span naming to use API names rather than model names to match JS implementation
• Added stop sequences attribute support and fixed handler removal bug

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
span_attributes.py	Renamed class from Span_Attributes to SpanAttributes and added GEN_AI_REQUEST_STOP_SEQUENCES attribute
callback_handler.py	Refactored parameter extraction into separate functions, updated span naming logic, and fixed attribute references

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

This code will raise a KeyError if 'kwargs' key doesn't exist in serialized dict. Should use serialized.get("kwargs") instead.

Suggested change

	if serialized and serialized["kwargs"]:
	if serialized and serialized.get("kwargs"):

[yiyuan-he]

I don't think we should go against the semantic conventions to align with JS.

Was there a reason we didn't align with semantic conventions in JS?

[yiyuan-he]

LGTM