Add Tempr's Mistral models

[chrissalomon]

Follow-up to #7638 (Tempr as a provider) and #7674 (Tempr's Google models): Tempr's Mistral models, 9 in total. Tempr addresses them as mistral/<id>.

Only models Mistral currently serves (checked against Mistral's model page on 2026-09-21): Mistral Large 3, Medium 3.5 and Small 4 (dated ids and -latest aliases), the GLM-5.2 and GLM-5.3 that Mistral hosts (base_model points at zhipuai/), and mistral-embed. Left out:

• models Mistral has retired (Mistral Nemo, Pixtral, Ministral 24.10, Mistral Large 2411, Medium 2505, Small 2506, Codestral 2501, Magistral, Voxtral Small, the open-mistral/open-mixtral models);
• codestral-latest and the ministral-*-latest aliases, whose catalog entries still describe the retired models they used to point at;
• the OCR, speech and moderation models, which no Tempr endpoint reaches.

The entries: base_model points at each lab entry with Mistral's own pricing, since it's the same model on the caller's own key. reasoning_options are what Tempr's GET /v1/models reports, and each file's leading comment lists the fields a caller sets. Reasoning docs: https://temprhq.io/docs/gateway-chat-completions#reasoning

Lab file: mistral-embed had no models/mistral/ entry, so this adds one, copied from providers/mistral without the pricing. Embeddings docs: https://temprhq.io/docs/gateway-embeddings

bun validate passes with this branch merged onto current dev.

🤖 Generated with Claude Code