docs: document LLMTR OpenAI-compatible setup #3372

Proposal

Add a short LLMTR section to docs/local-openai-compatible-providers.md, next to the OmniRoute one (#3283). Disclosure: I maintain LLMTR.

LLMTR is an OpenAI-compatible gateway at https://llmtr.com/v1 with Bearer keys (llmtr-...). Claw already works with it through OPENAI_BASE_URL and OPENAI_API_KEY; the docs just need to cover three things that trip people up:

• Some ids already start with the gateway name (llmtr/qwen3-8-27b). With local/llmtr/qwen3-8-27b, the local/ hatch is removed and the llmtr/ prefix is sent as-is.
• A bare qwen/... id is a DashScope route, so vendor-prefixed ids need local/ too.
• Claw asks for up to 64,000 output tokens on this route. LLMTR returns a 400 when max_tokens is above a model's output limit, so e.g. local/qwen/qwen3.8-27b-free (cap 16,384) fails today:

error: api returned 400 Bad Request (invalid_request): "qwen/qwen3.8-27b-free" can return at most 16,384 tokens in one response, but "max_tokens" was 64,000.

local/llmtr/qwen3-8-27b (262,144) works:

export OPENAI_BASE_URL="https://llmtr.com/v1"
export OPENAI_API_KEY="llmtr-..."
claw --model "local/llmtr/qwen3-8-27b" prompt "Reply exactly HELLO_WORLD_123"
# HELLO_WORLD_123

I have a docs-only PR ready and will link it here.