Add scipy optimisers

[MichaelClerx]

Nelder-Mead --> See #593

[MichaelClerx]

Wonder if we can get these in an ask-and-tell form. Could even consider re-implementing a few...

[MichaelClerx]

No derivatives:

• Nelder-mead (downhill simplex)
• Powell
• Differential evolution
• Basin hopping

First order derivatives:

• Conjugate gradient (cg) --> Does like a continuous gradient (and function)
• BFGS, Broyden, Fletcher, Goldfarb, and Shanno quasi-Newton method
• SLSQP, Sequential Least SQuares Programming

Second order derivatives:

• NCG Conjugated Newton or truncated Newton method --> Requires gradient, and (hessian OR hessian * arbitrary vector)

Constraining functions:

• Cobyla (Constrained Optimization BY Linear Approximation)