Linear Regression Prediction

Implement the prediction step and loss computation for a linear regression model.

prediction(x) = dot(w, x) + b

MSE loss = (1/n) * sum((y_pred - y_true)^2)

Given weights w, bias b, and features x, compute the prediction. Given a batch of predictions and true labels, compute MSE.

This is the forward pass of the simplest neural network -- a single neuron with no activation function.