Add Qwen model token limits for DashScope compatibility

[wangguan1995]

Summary

Fixes DashScope compatibility for Qwen models by adding explicit token limits for qwen-max and
qwen-plus, preventing claw from sending an oversized max_tokens value when routing through the
OpenAI-compatible provider.

Root Cause

rust/crates/api/src/providers/mod.rs only defines output token limits for a small set of known
models.
When qwen/qwen-max was used, the provider normalized it to qwen-max but did not find a
matching entry in model_token_limit(), so it fell back to the generic default in
max_tokens_for_model(). That default was larger than DashScope allows for qwen-max, causing
the API to reject requests with:

  400 Bad Request: Range of max_tokens should be [1, 8192]

Changes
Two small additions in rust/crates/api/src/providers/mod.rs:

• Add an explicit ModelTokenLimit entry for qwen-max
• Add an explicit ModelTokenLimit entry for qwen-plus

This ensures Qwen requests use a provider-specific max_tokens ceiling instead of the generic
fallback.

Verification

• cargo build --workspace succeeds
• ./target/debug/claw --model "qwen/qwen-max" prompt "hello" now succeeds
• Before the change, the same command failed with a DashScope max_tokens range error

Before / After

Before

./target/debug/claw --model "qwen/qwen-max" prompt "hello"
✘ Request failed
error: api returned 400 Bad Request
InternalError.Algo.InvalidParameter: Range of max_tokens should be [1, 8192]

After

./target/debug/claw --model "qwen/qwen-max" prompt "hello"
✔ Done
Hello! How can I assist you with your Rust project today?