fix: [missing-model] github-copilot: gemini-3.8-flash

[github-actions]

Summary

Adds the missing GitHub Copilot catalog entry for remote model gemini-3.8-flash. Lab metadata already exists at models/google/gemini-3.8-flash.toml; the provider file is override-only via base_model.

Changes

• New providers/github-copilot/models/gemini-3.8-flash.toml
  • base_model = "google/gemini-3.8-flash"
  • reasoning_options: effort low / medium / high
  • cost: input 0.75, output 3.75, cache_read 0.075 (USD / 1M tokens)
  • limit: context 1_000_000, input 936_000, output 64_000 (same Copilot host caps as gemini-3.7-flash)
  • modalities.input: ["text", "image"] (same Copilot host override as gemini-3.7-flash)

Evidence

Claim	Source	What it establishes
Model is available on GitHub Copilot as GA	Supported AI models	Lists Gemini 3.8 Flash (Google, GA)
Pricing $0.75 / $0.075 / $3.75 per 1M tokens	Models and pricing	Google table row for Gemini 3.8 Flash; promo note through 2026-12-31
Underlying lab model identity & reasoning	Existing models/google/gemini-3.8-flash.toml + first-party providers/google/models/gemini-3.8-flash.toml	Lab entry; first-party effort low/medium/high
Copilot host limits/modalities/effort shape	Existing providers/github-copilot/models/gemini-3.7-flash.toml	Same-family peer on this host: 1M context / 936k input / 64k output, text+image input, effort L/M/H

Validation

• Cross-checked GitHub Copilot pricing and supported-models docs against the new fields.
• Matched override pattern to gemini-3.7-flash on the same provider and lab/first-party effort set for Gemini 3.8 Flash.
• Did not run bun validate (shell disabled in this workflow).
• Did not call live https://api.githubcopilot.com/models (auth-gated).

Review notes

• Host classification: multi-model relay (github-copilot). Reasoning effort copied from lab first-party + same-host gemini-3.7-flash peer (not the older gemini-3.6-flash budget_tokens shape).
• Copilot docs confirm configurable reasoning for recent models but do not publish the exact effort enum; L/M/H is the established catalog baseline for this generation on Copilot and Google.
• Limit/modality overrides follow the Copilot gemini-3.7-flash peer rather than full lab multimodality (video/audio/pdf), consistent with how this host authors recent Gemini Flash entries.

Closes #6226

Automated by the issue fixer: https://github.com/anomalyco/models.dev/actions/runs/33807677890