Add inference loss

[DesmondYuan]

• Add a loss metric for parameter evaluation
• Record and plot both prediction performance and inference performance

[DesmondYuan]

Implemented in 09cc46afffc704b96a08ffa1a315517a37b9f499 0c00a66b74e5b6620aa2c9d61d2ef3bf65c09bc3 by @judyueshen

https://github.com/jiweiqi/CellBox.jl/blob/0c00a66b74e5b6620aa2c9d61d2ef3bf65c09bc3/spilt/network.jl#L29-L35

[jiweiqi]

It might be a good idea to add the scatter plot of p/w as well, something @judyueshen has done before.

[jiweiqi]

• We might plot w during the training as well.
• To do that, we might not want to generate a new figure for each iteration. So that we can simply use cbp(p, 0)

[jiweiqi]

This can be done by adding cbp(p, 0) at https://github.com/jiweiqi/CellBox.jl/blob/e9f112446718d81a9fd601335f7759c05bc6f20f/spilt/callback.jl#L53

[DesmondYuan]

Done here d619a2d7577bf9fff39398225667352e87e5d5ab