feat(core): add streaming output support for real-time token display

[appleboy]

• Add CompletionStream method to Generative interface with io.Writer
• Implement streaming in OpenAI provider using CreateChatCompletionStream
• Implement streaming in Anthropic provider using CreateMessagesStream
• Implement streaming in Gemini provider using GenerateContentStream
• Add --stream flag to commit and review commands bound to openai.stream
• Add callCompletion helper to route streaming or non-streaming calls
• Add streaming tests for OpenAI, Anthropic, and Gemini providers

[Copilot]

Pull request overview

Adds token streaming support across the core generative interface and all built-in providers, enabling real-time output in the CLI (commit/review) via a --stream flag.

Changes:

• Extends core.Generative with CompletionStream(ctx, content, w) and implements streaming in OpenAI, Anthropic, and Gemini providers.
• Adds a CLI --stream flag for commit and review, routing requests through a shared callCompletion helper.
• Introduces provider-level streaming tests (OpenAI/Anthropic SSE mocks; Gemini via mocked transport with conditional skip).

Reviewed changes

Copilot reviewed 10 out of 10 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
core/openai.go	Extends the core generative interface to support streaming output via io.Writer.
provider/openai/openai.go	Implements OpenAI streaming via CreateChatCompletionStream with usage capture.
provider/openai/stream_test.go	Adds an SSE-based test for OpenAI streaming accumulation + writer output.
provider/anthropic/anthropic.go	Implements Anthropic streaming via CreateMessagesStream callback.
provider/anthropic/stream_test.go	Adds an SSE-based test for Anthropic streaming accumulation + writer output.
provider/gemini/gemini.go	Implements Gemini streaming via GenerateContentStream with usage mapping.
provider/gemini/stream_test.go	Adds a mocked-transport streaming test (currently skips on error).
cmd/commit.go	Adds --stream flag and callCompletion helper to route streaming vs non-streaming.
cmd/review.go	Adds --stream flag and routes generation through callCompletion.
cmd/config_list.go	Documents openai.stream configuration key in config listing.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

callCompletion ignores the error from the trailing fmt.Fprintln(w) after streaming. If stdout/file is closed or fails, the command will still proceed as if successful; consider checking and returning this error.

Suggested change

	fmt.Fprintln(w)
	if _, err := fmt.Fprintln(w); err != nil {
	return resp, err
	}

[Copilot]

This test calls t.Skipf on any error from CompletionStream, which can mask real regressions (the test may silently become a no-op). Prefer asserting no error, or only skipping on a narrowly identified/expected SDK limitation error so failures still surface.

Suggested change

	t.Skipf("Skipping streaming test due to SDK transport constraints: %v", err)
	return
	}
	if resp.Content == "" {
	t.Error("expected non-empty content")
	}
	t.Fatalf("CompletionStream returned an unexpected error: %v", err)
	}
	if resp.Content == "" {
	t.Error("expected non-empty content")
	}