feat: predefined query for extracting tool calls

[tinuademargaret]

Issue: Fixes #7158

Description:
This PR introduces a predefined query get_called_tools for querying LLM tool calls and their associated input/output messages from span data.

Example usage:

from phoenix.trace.dsl.helpers import get_called_tools

tools_df = get_called_tools(client)

# Example output:
# context.span_id  | tool_call
# ----------------|------------------
# span_1          | ["multiply", "add"]  
# span_2          | ["add"]            

This has been tested to work with the top llm providers and framework auto-instrumentations noted in the issue

[github-actions]

CLA Assistant Lite bot All contributors have signed the CLA ✍️ ✅

[tinuademargaret]

I have read the CLA Document and I hereby sign the CLA

[tinuademargaret]

I was wondering if the parameters of the tool should also be returned by the query, maybe have that as an option?

[Jgilhuly]

Thank you for this @tinuademargaret! Doing some final testing today with a range of providers

[Jgilhuly]

@RogerHYang - tested from my side with our top 10 providers. All working fine!

One note here: the helper exports the user input under the column header input, but our tool calling template looks for question as the header. We should be able to immediately put the output of this helper into the template, but I'd actually argue that input is the better header name. I'd rather we update the template to match this, than the other way around. Lmk your thoughts there

[RogerHYang]

Thank you for your contribution, @tinuademargaret!

I've added a flag to include function arguments to the result by default.