Add mixin class and inherit it in torchplate.experiment.Experiment

[rosikand]

Add a bunch of experiment utility functions in a mixin class to be inherited in torchplate.experiment.Experiment (and therefore also inherited in subclasses of torchplate.experiment.Experiment). Example methods to include:

• test_model_accuracy for classification experiments

[rosikand]

Bumping this to include a simple test loop for a classification model next update. Should be of the form:

    def test(self, model, testloader, weights):
        # if weights is not None..., load weights 
        acc = tp_metrics.Accuracy()
        tqdm_loader = tqdm(testloader)
        for batch in tqdm_loader:
            x, y = batch
            logits = model(x)
            acc.update(logits, y)
            tqdm_loader.set_description(f'Accuracy: {acc.get()}')
        
        final_acc = acc.get()
        print(f'Accuracy: {final_acc}')
        acc.reset()

        return final_acc

can even include a separate version that uses self. attributes.