chore(weave): update model providers

[mattliu-mygit]

What

Regenerate model_providers.json from the current W&B hosted-model catalog and LiteLLM model catalog. The result has 1,903 routes: 256 additions, 176 removals, and one provider change for computer-use-preview from Azure to OpenAI.

This PR supersedes #7917, which was generated from the September 21 catalog snapshot.

Why

The checked-in routing catalog is stale. LiteLLM added current model and regional routes, then removed models past their deprecation date and models flagged as retired by provider sync.

How

Run make update_model_providers without hand-editing the generated file. The source snapshot audited for this refresh is LiteLLM commit 58763b3021.

The generated keyset matches supported providers in that snapshot plus W&B-hosted routes. The generator also preserves 23 existing deprecation annotations for routes that remain present.

Testing

• Reran make update_model_providers and confirmed the generated SHA-256 hash was unchanged.
• Parsed the generated JSON and compared every route's provider and API-key mapping with the two source catalogs.
• Confirmed git diff --check passes.

Anything Else

• This PR changes routing metadata only. It does not add models to a playground dropdown.
• The 176 removals trace to LiteLLM's explicit retired-model cleanup commits #42435 and #42521.
• Cost data is regenerated in a separate PR so routing and pricing remain independently reviewable.

[ro31337]

@mattliu-mygit Heads-up: we're moving the trace server (weave/trace_server/ and its tests) to wandb/core. These paths will soon be frozen and then removed from this repo, so this PR may need to move to core. Please follow #weave-instrumentation on Slack, and ask there if you have questions. If this PR is urgent, reach out in the channel early so we can sync, or wait for the README.md update.