Assuming that the speed of the toy car is constant, which may not always be the case, the uncertainty here will be dominated by your reaction time and your visual perception. I would not worry about the ruler and the stopwatch.
Making a bunch of measurements and averaging the results will reduce your random error. Extending the length of the track will reduce both random and systematic error, since your reaction time will have a smaller contribution to the measured time.
But, let's say, you want to estimate the uncertainty of the calculated speed based on a single measurement, performed as you've described, and, for that, you need to know the uncertainties of your measurements. One possible approach here is to perform some preliminary tests to "calibrate" yourself, so to speak, since, as mentioned before, your reaction time and visual perception will be dominant factors affecting the uncertainty of your results.
You can estimate your reaction time by timing known intervals with your stopwatch. Here is a reference clock you can use for this purpose. You can make a number of measurements of some interval, say, $3$ seconds, and determine the standard deviation of your measurements, which will give you an idea of uncertainty in the measured time.
To estimate the uncertainty in the perception of distance, you can place the toy car at random locations close to the beginning and then close to the end of the meter stick and try to quickly assess the position of the car from the location(s) you use to perform your speed measurements. By doing it a number of times and then comparing your quick assessments with more carefully measured positions of the car, you'll get an idea of the uncertainty in the measured distance.