### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

When using Bedrock prompt caching with a tool-level cache injection point (cache_control_injection_points with location: "tool_config", see [LiteLLM docs — Anthropic Cache Control (Bedrock)](vscode-file://vscode-app/Applications/Visual%20Studio%20Code.app/Contents/Resources/app/out/vs/code/electron-browser/workbench/workbench.html)), the cache marker is correctly sent to Bedrock — confirmed by real [cache_creation_input_tokens](vscode-file://vscode-app/Applications/Visual%20Studio%20Code.app/Contents/Resources/app/out/vs/code/electron-browser/workbench/workbench.html) and [cache_read_input_tokens](vscode-file://vscode-app/Applications/Visual%20Studio%20Code.app/Contents/Resources/app/out/vs/code/electron-browser/workbench/workbench.html) values in the API response, and this matches Bedrock's documented request format for caching tool definitions (see [AWS Bedrock — Prompt caching, "tools checkpoints" example](vscode-file://vscode-app/Applications/Visual%20Studio%20Code.app/Contents/Resources/app/out/vs/code/electron-browser/workbench/workbench.…