fix(provider): add Standard Compute as a built-in provider

[henrikdukefoss]

Adds Standard Compute as a built-in provider. I work on Standard Compute.

Standard Compute is an OpenAI-compatible gateway: one key for models from several labs, on monthly plans with a fixed compute budget. The entry uses the OpenAI response type and the existing request path, so no new transport code is needed. It follows the OrcaRouter and Ambient entries.

Changes

• crates/forge_repo/src/provider/provider.json: standard_compute entry with STANDARDCOMPUTE_API_KEY, OpenAI response type, chat URL https://api.stdcmpt.com/v1/chat/completions and models URL https://api.stdcmpt.com/v1/models, api_key auth.
• crates/forge_domain/src/provider.rs: ProviderId::STANDARD_COMPUTE, added to built_in_providers(), display_name and FromStr. Three tests mirror test_orca_router_*, and the id joins the two shared assertion lists.
• crates/forge_repo/src/provider/provider_repo.rs: test_standard_compute_config, mirroring test_orca_router_config (including the Models::Url check).
• README.md: env-var block next to OrcaRouter.

Models

The model list is fetched from the public /v1/models endpoint (no key needed), which returns the plain OpenAI list shape that ListModelResponse already parses, with context_length, architecture.input_modalities and supported_parameters. standardcompute routes each request to one of the models enabled in the account; exact IDs such as anthropic/claude-sonnet-5.5 pin a single model.

API docs: https://docs.standardcompute.com/api/chat-completions

Verification

• cargo test -p forge_domain --lib provider: 53 passed
• cargo test -p forge_repo --lib provider: 285 passed, 1 ignored (includes test_standard_compute_config)
• cargo clippy -p forge_domain -p forge_repo --tests: no warnings in the touched code
• cargo fmt --all -- --check: clean
• GET https://api.stdcmpt.com/v1/models returns {"object":"list","data":[...]} with the fields above

🤖 Generated with Claude Code

[CLAassistant]

All committers have signed the CLA.