Implement SGD under Algorithms/GradientOptimizers.hpp

[JackHunt]

Is your feature request related to a problem? Please describe.
Currently there is only a base optimiser class, there should be at least one concrete optimiser.

Describe the solution you'd like
An implementation (CPU only as a start) of simple SGD (no momentum, nester etc) following the API of GradientOptimizerBase. Simple test cases testing convergence and documentation should also be present.

Describe alternatives you've considered
Removing GradientOptimizerBase until a concrete optimiser is required, but this seems wasteful.

Additional context
Should not contain f and df implementations, only implementing the logic of the optimiser. The mathematical functions should be user provided.