fix(anthropic): add instrumentation for Anthropic tool calling (alternative to #1372)

[dinmukhamedm]

This commit addresses [#879] by implementing instrumentation of Anthropic tools request args and tool_use response arguments.

• I have added tests that cover my changes.
• If adding a new instrumentation or changing an existing one, I've added screenshots from some observability platform showing the change.
• PR name follows conventional commits format: feat(instrumentation): ... or fix(instrumentation): ....
• (If applicable) I have updated the documentation accordingly.

[dinmukhamedm]

@nirga looking for your input on this. I tried to stay as close to OpenAI as possible. For OpenAI the index, is the ordinal number of the output in the choices array. I believe, there is no such thing in Anthropic, so index is always 0.

In theory, Anthropic could respond with something like [text, tool_use, text, text, tool_use]. I've never seen it do this in practice, but just in case, instead of hard-coding, I decided to concatenate everything in the text block.

An alternative to my approach would be to set these text blocks as gen_ai.completions.i.content, but I don't think this is consistent with the choices in OpenAI.

[nirga]

Makes sense @dinmukhamedm! Long term we plan to move these to otel events so there will be a "cleaner" solution

[CLAassistant]

All committers have signed the CLA.

[dinmukhamedm]

Sorry about the back and forth, had some trouble running nx lint on my machine. Perhaps a good sign to add an instruction note to the contribution guideline? :)