Standardize output and add accuracy numbers

[rdadolf]

#13 helped, but the output is still a little messy to use. Tracking a loss curve is a very useful thing, but in practice, loss isn't sufficient a lot of times, and it can be confusing to non-DL people.

Updates to verbose output:

• standardized across all 8 models
• include both loss and accuracy
• scriptable output (cut -d ' ' -f x,y should basically do the right thing)
• don't impact performance measurements if possible (so try to avoid adding graph nodes)