feat(forge_main,forge_domain,forge_api,forge_app,forge_tracker): graceful provider model fetch

[KooshaPari]

Summary

Previously, a single broken provider (auth failure, network error, malformed response) would abort /model enumeration and lock the user out of switching models on healthy providers. This change introduces partial-success semantics so that:

• Successful providers continue to surface their models.
• Failed providers are reported inline with their per-provider error message.
• The aggregated ProviderModelsResult carries both halves so the UI can render /models with warnings instead of aborting.

Files touched

• forge_domain/src/provider.rs — adds ProviderFetchError + ProviderModelsResult
• forge_api/src/api.rs — threads the aggregated result through
• forge_api/src/forge_api.rs — small type adjustment
• forge_app/src/app.rs — collects per-provider failures
• forge_main/src/ui.rs — renders partial-failure warnings
• forge_main/src/update.rs — propagates through state machine
• forge_tracker/src/log.rs — structured log for failed fetches

Test plan

• Local build succeeds
• Cargo clippy clean

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}