Update X and Y during init

[ajnisbet]

I often plot/monitor the optimisation each iteration in case I want to stop early or change some parameters. I do this by adding monitoring code to the target function.

Currently, the x and y values checked during explore and init are not accessible (stored on the bo object) until after both steps are complete, which can take a long time for expensive functions.

This PR adds updates self.X and self.Y each iteration of the explore/init step. Most of the logic is converting append and += to numpy equivalents. Adds a bit of complexity but doesn't change any other part of the API.

You could go all out and also update bo.go.fit() and bo.res each iteration, and add callback parameter
to be called after each iteration to avoid hacking the target function like xgboost and some sklearn functions. But just updating X and Y is the minimum info needed for a user to just do that themselves.

[fmfn]

Looks good. Did you try running the usage.py script?

[ajnisbet]

Yup, same logs and solution.