fix(web_fetch): increase read limit and add empty content detection

- Read limit: maxChars*4 → max(maxChars*10, 512KB) to handle pages with
  large <head> sections (WordPress sites often have 30-50KB+ heads)
- Add warning message when HTML extraction returns empty despite non-empty
  response body (bot protection, JS-only pages)
- Enable HTTP/2 via ForceAttemptHTTP2 on custom Transport

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

internal/tools/web_fetch.go

@@ -140,6 +140,7 @@ func (t *WebFetchTool) doFetch(ctx context.Context, rawURL, extractMode string,
 	client := &http.Client{
 		Timeout: time.Duration(fetchTimeoutSeconds) * time.Second,
 		Transport: &http.Transport{
+			ForceAttemptHTTP2:   true,
 			MaxIdleConns:        10,
 			IdleConnTimeout:     30 * time.Second,
 			TLSHandshakeTimeout: 15 * time.Second,
@@ -163,8 +164,9 @@ func (t *WebFetchTool) doFetch(ctx context.Context, rawURL, extractMode string,
 	}
 	defer resp.Body.Close()
 
-	// Limit body reading to avoid memory issues
-	limitReader := io.LimitReader(resp.Body, int64(maxChars*4)) // read extra for HTML overhead
+	// Read enough HTML to reach <body> content — pages often have 30-50KB+ <head> sections.
+	readLimit := int64(max(maxChars*10, 512*1024))
+	limitReader := io.LimitReader(resp.Body, readLimit)
 	body, err := io.ReadAll(limitReader)
 	if err != nil {
 		return "", fmt.Errorf("read body: %w", err)
@@ -196,6 +198,10 @@ func (t *WebFetchTool) doFetch(ctx context.Context, rawURL, extractMode string,
 			text = htmlToText(string(body))
 			extractor = "html-to-text"
 		}
+		if text == "" && len(body) > 0 {
+			text = "[No content extracted. The page may require JavaScript to render, " +
+				"or returned a bot-protection challenge. Try using browser automation instead.]"
+		}
 
 	default:
 		text = string(body)