There are currently two methods of reporting errors in conkeror; the message area and the console.  The
message area is used to report mistakes in the user's interaction and other problems that conkeror
specifically notices.

Errors in conkeror itself and errors in web page content are both reported on the console.  That output
can be swamped by errors in web pages that the user does not control, masking errors in conkeror.  It may
be worth providing more control over where errors appear and how they are presented.

Some suggestions:

  * Put content errors in a log file, separate from conkeror errors.

  * Have much terser errors on the output.  Put details in a log file.

  * A *Messages* buffer.

  * User-configurable streams for different log levels.

Another possibility that some people might like is colorizing terminal output with ansi escape
codes.

I have in mind something that looks like this:

  var log_routes = {
      interactive_error: show_in_minibuffer,
      error: dumpln,
      warning: dumpln
  };

  function log (type, text, context) {
      var logfn = log_routes[type] || dumpln;
      logfn(text, context);
  }

Any given function in log_routes receives both the text and a context object.  The context object would
include properties like "buffer" and "window" when relevant.  This means that the interactive context,
"I", could be used also as a log context.