Feature: Support Prompt Caching

[kowshik24]

trafficstars

This PR supports tracking and logging prompt caching(#1838) metrics in the OpenTelemetry Anthropic instrumentation.

Specifically, it captures and logs the cache_creation_input_tokens and cache_read_input_tokens fields from the API response. This enhancement allows users to monitor the effectiveness of their prompt caching strategy, providing insights into cache utilization and efficiency.

Changes include:

• Updated _set_token_usage and _aset_token_usage functions to log caching metrics.
• Added new span attributes for cache_creation_input_tokens and cache_read_input_tokens.
• Updated streaming.py to support prompt caching metrics in streaming responses.

[nirga]

@kowshik24 note that you still have lint / test errors. LMK once you want me to re-review

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
1 out of 2 committers have signed the CLA.

:white_check_mark: kowshik24
:x: kowshik

---

kowshik seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[dinmukhamedm]

@nirga @kowshik24 What is the status of the changes here? If this is stale, we are happy to contribute!

[nirga]

Would love for you to contribute @dinmukhamedm! I think @kowshik24 is no longer working on this

[dinmukhamedm]

@nirga thanks, and sorry about the delayed response. This is requiring a little bit more work with counting tokens (and testing that) than I've expected, and I was out for a day. I'm still on this though