Fix usage tracking for SiliconFlow etc

[mrubens]

Context

As described in #1227, some APIs return the usage counts in every chunk, causing our current method of tracking to show completely inaccurate counts (and likely confusing the sliding window calculations).

Implementation

@tbphp had a great suggestion here to just use the last usage if there are multiple: #1227 (comment)

Screenshots

before	after

How to Test

I signed up for the SiliconFlow provider at https://siliconflow.cn/ and used the free deepseek-ai/DeepSeek-R1-Distill-Qwen-7B model to test.

---

Important

Fixes usage tracking in OpenAiHandler by using only the last usage metrics from API streams, with comprehensive tests added.

• Behavior:
  • Fixes usage tracking in OpenAiHandler by using only the last usage metrics from API streams.
  • Handles cases where usage data is only in the final chunk or not provided at all.
• Testing:
  • Adds openai-usage-tracking.test.ts to test usage metrics handling with multiple chunks, final chunk only, and no usage data scenarios.
• Implementation:
  • Modifies createMessage() in openai.ts to store and yield the last usage metrics.

^{This description was created by for f306461. It will automatically update as commits are pushed.}

[changeset-bot]

🦋 Changeset detected

Latest commit: f306461

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
roo-cline	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[hannesrudolph]

I swear we had a PR or two handling this.

[mrubens]

Yeah #962 was another, I think we can close it now