LBFGS Support

[joverwey]

Is there a reason why LBFGS is not supported as an optimizer? Would it be hard to add?

For Style Transfer, I've noticed that it converges much faster than AdamW when used from Python.

Also see this comparison with other optimizers: https://blog.slavv.com/picking-an-optimizer-for-style-transfer-86e7b8cba84b