chore: Adjust llm.tools semantic convention example

[nate-mar]

Adjusting the example for llm.tools to make it consistent with how other list of object semantic conventions are repented in the documentation, so that it doesn't convey that there's something actually different in its structures compared to any other list of objects semantic convention, such as llm.input_messages and llm.output_messages. Ultimately, these open inference semantic conventions are flattened and not nested when ingested via OTLP (e.g. llm.tools.0.tool.id, llm.tools.1.tool.id` for semantic conventions that use list of objects), so there is no actual difference in the structure of the data between semantic conventions of the same type.

Furthermore, some systems may choose to represent conventions like a list of objects as a single column as opposed to a flattened set of values, so this doc change also ensures that users know they can expect to consistently handle conventions of the same type (e.g. list of objects).