[Feat] : Add weight decay support to all optimizers

[Advaitgaur004]

Added L2 regularisation (weight decay) to SGD, AdaGrad, RMSProp, and Adam optimisers. Each optimizer now takes an additional weight_decay parameter in its constructor.

• Modified all optimiser constructors to accept a weight decay parameter
• Gradient calculation now includes the weight_decay * param_value term

Quick Eg.

// Before: optim_sgd_new(n_params, params)
// After: optim_sgd_new(n_params, params, 0.01f)  // 0.01 weight decay

Should help keep those models from going crazy on the training data!!! 😮‍💨