Allow switching AI model provider after init without overwriting existing files

[ArpitSuthar]

Describe the problem
After running specify init , changing the configured AI model provider requires re-running init or manual edits. Re-running init can overwrite existing project files (scaffolding, constitution.md, .specify, prompts, or custom templates), risking loss of progress made so far.

Proposed solution
Add a provider-switch command or enhance init to support switching providers safely by:

• Introducing a command like specify provider switch <provider-name> (or specify switch-provider) that updates provider-specific configuration only.
• Detecting existing files and merging provider-specific changes instead of replacing files. Use a safe merge strategy:
  • Update only provider-specific config entries in .specify or equivalent config file.
  • Preserve custom files (constitution.md, prompts/, templates/) unless a --force flag is provided.
  • When provider-specific template files are missing, create provider variants rather than overwriting the generic ones.
• Add an interactive prompt/confirmation showing which files would be changed and an explicit --force option to allow overwriting when desired.
• Provide a dry-run mode (--dry-run) that shows diffs/changes.

Additional context / notes

• Consider supporting provider-specific prompt/template directories to allow coexistence (e.g., .openai/, .anthropic/).
• Ensure migration path for projects that used older init flows—document steps to move provider-specific files into the new layout.

[dominikj-cf]

Supporting multiple providers is pretty fundamental in larger teams that may have access to different preferred CLI tools.

[mnriem]

Delivered by #2083