[Feature]: Text-generation failure popups should name the attempted model and link to Settings

Context

#12651 confirmed that one-shot text generation (commit messages, PR content, branch names, thread titles) can fail opaquely: the Commit popup showed an unrelated stderr notice while the real cause was an unconfigured/unsupported text-generation model. #12652 improves what the error says (it now surfaces the CLI's own stdout error), but the popup still leaves the user to figure out where the fix lives.

Proposal

When one-shot text generation fails, make the error actionable by surfacing what the server already knows and offering the fix in place:

1. Name the attempted model and its origin. E.g. if textGenerationModelSelection is unset and a fallback was used, say so explicitly: "No text generation model was configured, so the provider default (claude-haiku-4-5) was used." That is server-known information — resolveTextGenerationProvider() knows whether it fell back and to which slug.
2. Include a button/link that opens Settings at the text-generation model section. One click from the Commit popup to the place where the model is configured.
3. Same treatment for title/branch failures, which today can fail silently (#5359) — a consistent actionable-failure path would cover the whole family.

Why

The failure mode in #12651 would have been a two-click self-fix from the Commit dialog itself: popup names the fallback model → button opens Settings → user picks a model their backend allows. Instead it took a full diagnostic session (captured spawns, extracted bundle source) to discover that textGenerationModelSelection was unset and that the fallback slug was being rejected by the backend.

The information needed for this already exists at the failure site; this is purely about presenting it.

Impact

Quality-of-life enhancement on top of the #12651 fixes.

[!NOTE]
🤖 GLM 5.3 Flash on behalf of iliasgws

Inspecciona el popup de Commit y la ruta resolveTextGenerationProvider(), usando #12651 y #12652 para el comportamiento de error existente. Se considera terminado cuando los errores de generación de una sola ejecución identifican el modelo y el origen intentados, proporcionan un enlace a la sección de modelos de generación de texto en Settings y aplican el mismo flujo accionable a los errores de título y de branch.