Classification: Layer error: Classifier training failed: 'Only one class.'

var classifier = ee.Classifier.smileCart().train(training, 'landcover', bands);

You're telling the classifier to learn to classify points according to the value of the 'landcover' property. But in all of your input data, the value of landcover is always 0, so there is only one class, 0. This is reported as an error because a classification with one class is not useful — it gives no information.

To fix this, in the geometry list, click on the gear icon for each of your point sets:

Then in the "Configure geometry imports" dialog, give each one a different number for the landcover property. I edited them so that urban is 0, forest is 1, agriculture is 2, and water is 3.

After you do this, the script will run without error and your classifier will be trained successfully. Then you'll want to view the results. Here's how to do that with a palette set up to match your marker colors:

Map.addLayer(classified, {min: 0, max: 3, palette: ['red', 'blue', 'green', 'cyan']});

Here is your script with those two modifications (property value in the geometry imports, and Map.addLayer in use).