Support custom VS Code chat model providers for generation and autopilot

[Copilot]

Spec generation and autopilot were hard-wired to Copilot gpt-4o, which prevented the extension from working with custom chat models exposed by other VS Code providers. This change makes model selection provider-agnostic while preserving the current default preference when Copilot is available.

• Model selection

  • added a shared language-model selection path for spec generation and autopilot
  • introduced copilot-specs.languageModelSelector to target models by id, vendor, family, or version
  • kept the default behavior of preferring Copilot gpt-4o when no selector is configured, with fallback to any available VS Code chat model

• Selector semantics

  • normalized selector parsing into a dedicated pure helper
  • matched configured selectors strictly across all provided fields instead of silently drifting to unrelated models
  • separated pure matching logic from VS Code runtime integration to keep selection behavior explicit and testable

• User-facing behavior

  • removed Copilot-only availability messaging from generation/autopilot flows
  • updated prompts/errors to refer to compatible VS Code chat models rather than only GitHub Copilot

• Configuration and docs

  • added the new setting to package.json
  • documented custom model selection in README.md
  • recorded the behavioral change in CHANGELOG.md

Example configuration:

"copilot-specs.languageModelSelector": {
  "vendor": "my-provider",
  "family": "llama",
  "version": "1.0"
}