Copilot CLI integration follow-up and private-network local LLM support

[provenvelocity]

Thanks for the initial Copilot CLI adapter in #15881! I used that PR as the starting point for additional integration work in my fork: https://github.com/provenvelocity/paperclip.

The work implemented and tested locally includes:

• Merged newer upstream changes into the Copilot contribution while preserving the original author's commits.
• Added dashboard-managed GitHub token support for Copilot CLI agents.
• Improved token permission guidance and connection verification using a direct CLI hello probe.
• Added regression coverage for saved managed connections.
• Added purely local, no-key OpenAI-compatible endpoint support over HTTP on private networks, rather than requiring loopback or a public cloud provider.
• Added optional certificate trust for self-signed HTTPS on private endpoints, without disabling TLS verification globally.
• Restricted HTTP and custom-certificate exceptions to private IPs and loopback; public HTTP remains rejected.
• Fixed the New Agent dialog crashing on plain-HTTP LAN deployments.
• Built and deployed the changes locally, with targeted tests and deployment checks.

Scope clarification: private local-endpoint support is implemented in Paperclip's model-connection layer. Wiring those managed connections into the Copilot adapter's custom-provider launch environment remains separate work; this implementation does not already complete that integration. Private-endpoint validation is also not a general egress firewall.

These additional changes are currently local and are not yet published to the linked fork. Targeted checks passed, but the full-suite baseline still has upstream failures, so I am not claiming a fully green, PR-ready branch.

Would the maintainers and original author be interested in having me publish a focused branch and contribute an update to #15881? I can separate the Copilot-specific improvements from the broader private-network endpoint support if that would make review easier, and follow whichever contribution path you prefer.