Custom OpenAI-compatible providers: add per-model display_name, reasoning, max_output_tokens, cost, and provider display_name (opencode parity)

Summary

Custom OpenAI-compatible providers configured via [providers.<name>] in ~/.jcode/config.toml were missing several fields that opencode supports for its custom OpenAI-compatible provider spec. This made it harder to get a clean model picker experience and correct cost display for self-hosted or third-party gateways.

Gaps identified

Changes

Config schema (jcode-config-types)

• New ModelCostConfig struct: input, output, cache_read, cache_write (USD per million tokens)
• New fields on NamedProviderModelConfig: display_name (alias name), max_output_tokens (alias output_limit), reasoning (bool), cost (ModelCostConfig)
• New field on NamedProviderConfig: display_name (provider-level)

Runtime wiring (jcode-provider-openrouter-runtime)

• available_models_display(): applies per-model display name aliases in the picker
• runtime_display_name() + direct_openai_compatible_route_parts(): use provider display_name
• supports_deepseek_reasoning_effort() / supports_openai_reasoning_effort(): check per-model reasoning override first
• complete(): per-model max_output_tokens overrides runtime default in the request body
• fork(): carries all new fields

Picker routes (jcode-base/provider/catalog_routes.rs)

• named_provider_profile_routes(): uses provider display_name as route label, per-model display_name as route model field, and converts cost config into RouteCheapnessEstimate for the picker cost overlay
• New cost_to_route_cheapness() helper

Tests

• 7 new tests in openrouter_tests.rs covering display name, reasoning, max_output_tokens, and cost
• 2 new tests in catalog_routes.rs covering route display name + cost cheapness conversion

Docs

• README updated with new config field descriptions and enriched config.toml example

Config example

[providers.my-api]
type = "openai-compatible"
base_url = "https://llm.example.com/v1"
api_key_env = "JCODE_PROVIDER_MY_API_API_KEY"
default_model = "my-model-id"
display_name = "My Custom API"

[[providers.my-api.models]]
id = "my-model-id"
context_window = 128000
display_name = "My Model"
max_output_tokens = 8192
reasoning = true
cost = { input = 0.15, output = 0.60, cache_read = 0.015 }

Branch

fix/custom_openai_compatible

Verification

• Full workspace cargo check passes cleanly
• 99 jcode-provider-openrouter-runtime lib tests pass (7 new)
• 2 jcode-base catalog_routes tests pass (2 new)
• All existing tests unaffected

Start with the config types in jcode-config-types, then trace provider handling through jcode-provider-openrouter-runtime and jcode-base/provider/catalog_routes.rs. Run the named openrouter_tests.rs and catalog_routes.rs tests; done means the documented provider and model fields are represented in runtime requests and picker routes, with the README example consistent.