feat: Add one-off prompt command with GitHub issue integration

[edenreich]

Implements background mode execution for prompts that can:

• Parse GitHub issue references in multiple formats
• Fetch GitHub issue context using GitHub API
• Execute prompts iteratively until task completion
• Support up to 10 iterations with completion detection

Closes #22

Generated with Claude Code

[edenreich]

@claude Please fix the CI - I've fast-forward this branch from main upstream.

[claude]

Claude finished @edenreich's task —— View job

---

CI Issues Fixed! ✅

Successfully resolved the CI type error and committed the fix to the PR branch.

Todo List:

• Check CI status and identify failures
• Set up development environment
• Examine the type error in cmd/prompt.go:294
• Fix the type error using the FormatToolResultForLLM function
• Run local quality checks to verify fix
• Commit fixes to the PR branch

Issue Fixed: The code was trying to return a *domain.ToolExecutionResult as a string. I added the missing ui import and changed the return statement to use ui.FormatToolResultForLLM(result) to properly convert the result to a string.

Changes Made:

• Added import for "github.com/inference-gateway/cli/internal/ui"
• Changed return result, nil to return ui.FormatToolResultForLLM(result), nil

Verification:

• ✅ Linter passes (0 issues)
• ✅ Tests pass
• ✅ Build succeeds

Commit: ba6c21c - "fix: Resolve type error in executeToolCall function"

The CI should now pass on the next run. The fix ensures that tool execution results are properly formatted as strings for LLM consumption, which maintains the expected interface contract.