refactor api package comments

* use encoding/json for comment parsing

* refactor by moving loop code to an UnmarshalJSON

* use a preallocated array and indices to maintain order while using
  goroutines again, this was removed a while ago

* use new struct in comment.hbs and contextComment.hbs

* rewriteUrl partial to reduce rimgo-specific code in api

* move RenderError into pages package to avoid import cycle between render and utils

pages/error.go

@@ -0,0 +1,44 @@
+package pages
+
+import (
+	"fmt"
+	"io"
+	"net/http"
+	"strconv"
+
+	"codeberg.org/rimgo/rimgo/render"
+	"codeberg.org/rimgo/rimgo/static"
+	"codeberg.org/rimgo/rimgo/utils"
+)
+
+func RenderError(w http.ResponseWriter, r *http.Request, code int, str ...string) (err error) {
+	if len(str) != 1 {
+		str = []string{""}
+	}
+	codeStr := "generic"
+	if code == 0 {
+		code = 500
+	}
+	if code != 500 {
+		codeStr = strconv.Itoa(code)
+	}
+	if !utils.Accepts(r, "text/html") && r.PathValue("extension") != "" {
+		w.Header().Set("Content-Type", "image/png")
+		w.WriteHeader(code)
+		file, _ := static.GetFiles().Open("img/error-" + codeStr + ".png")
+		defer file.Close()
+		_, err = io.Copy(w, file)
+
+	} else {
+		w.WriteHeader(code)
+		err = render.Render(w, "errors/"+codeStr, map[string]any{
+			"path": r.URL.Path,
+			"err":  str[0],
+		})
+	}
+	if err != nil {
+		// don't panic or return error, it will loop
+		fmt.Println("error in RenderError: " + err.Error())
+	}
+	return nil
+}