feat: comprehensive token usage tracking for V2

Add provider-specific usage details, fix streaming usage, and return
usage from all high-level APIs (Chat.Send, Generate[T], Agent.Run).

Breaking changes:
- Chat.Send/SendMessage/SendWithImages now return (string, *Usage, error)
- Generate[T]/GenerateWith[T] now return (T, *Usage, error)
- Agent.Run/RunMessages now return (string, *Usage, error)

New features:
- Usage.Details map for provider-specific token breakdowns
  (reasoning, cached, audio, thoughts tokens)
- OpenAI streaming now captures usage via StreamOptions.IncludeUsage
- Google streaming now captures UsageMetadata from final chunk
- UsageTracker.Details() for accumulated detail totals
- ModelPricing and PricingRegistry for cost computation

Closes #2

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

v2/google/google.go

@@ -59,12 +59,32 @@ func (p *Provider) Stream(ctx context.Context, req provider.Request, events chan
 
 	var fullText strings.Builder
 	var toolCalls []provider.ToolCall
+	var usage *provider.Usage
 
 	for resp, err := range cl.Models.GenerateContentStream(ctx, req.Model, contents, cfg) {
 		if err != nil {
 			return fmt.Errorf("google stream error: %w", err)
 		}
 
+		// Track usage from the last chunk (final chunk has cumulative counts)
+		if resp.UsageMetadata != nil {
+			usage = &provider.Usage{
+				InputTokens:  int(resp.UsageMetadata.PromptTokenCount),
+				OutputTokens: int(resp.UsageMetadata.CandidatesTokenCount),
+				TotalTokens:  int(resp.UsageMetadata.TotalTokenCount),
+			}
+			details := map[string]int{}
+			if resp.UsageMetadata.CachedContentTokenCount > 0 {
+				details[provider.UsageDetailCachedInputTokens] = int(resp.UsageMetadata.CachedContentTokenCount)
+			}
+			if resp.UsageMetadata.ThoughtsTokenCount > 0 {
+				details[provider.UsageDetailThoughtsTokens] = int(resp.UsageMetadata.ThoughtsTokenCount)
+			}
+			if len(details) > 0 {
+				usage.Details = details
+			}
+		}
+
 		for _, c := range resp.Candidates {
 			if c.Content == nil {
 				continue
@@ -105,6 +125,7 @@ func (p *Provider) Stream(ctx context.Context, req provider.Request, events chan
 		Response: &provider.Response{
 			Text:      fullText.String(),
 			ToolCalls: toolCalls,
+			Usage:     usage,
 		},
 	}
 
@@ -284,6 +305,16 @@ func (p *Provider) convertResponse(resp *genai.GenerateContentResponse) (provide
 			OutputTokens: int(resp.UsageMetadata.CandidatesTokenCount),
 			TotalTokens:  int(resp.UsageMetadata.TotalTokenCount),
 		}
+		details := map[string]int{}
+		if resp.UsageMetadata.CachedContentTokenCount > 0 {
+			details[provider.UsageDetailCachedInputTokens] = int(resp.UsageMetadata.CachedContentTokenCount)
+		}
+		if resp.UsageMetadata.ThoughtsTokenCount > 0 {
+			details[provider.UsageDetailThoughtsTokens] = int(resp.UsageMetadata.ThoughtsTokenCount)
+		}
+		if len(details) > 0 {
+			res.Usage.Details = details
+		}
 	}
 
 	return res, nil