Add GitHub Copilot as an inference backend

[sgoedecke]

Summary

• add --provider so gh standup can use either GitHub Models or GitHub Copilot
• wire the Go Copilot SDK into the LLM client while keeping GitHub Models as the default path
• document Copilot CLI setup and add tests for provider/model selection helpers

Notes

• Copilot runs through the official Copilot SDK / Copilot CLI path
• the Copilot SDK requires Go 1.24, so this PR updates the module Go version accordingly
• Copilot sessions use a deny-all permission handler because this command only needs inference, not tool execution