Bug with loss parsing

[motiwari]

https://github.com/ThrunGroup/BanditPAM/blob/main/src/main.cpp#L40-L42 contains a bug; it assumes that the single character after the L is the norm; this creates an error, for example, when the loss is L15, or if the name of a loss function happens to start with L.

Instead, we should check specifically for the form L\d* and get all characters except the first in that case.