Per-label and per-class metrics should be formatted properly for CSV output

[sarthakpati]

Describe the bug
The per class/label metrics get saved as lists in the CSV, which makes parsing difficult.

To Reproduce
Steps to reproduce the behavior:

1. Run any training.
2. See the following in logs (example shown for a simple classification unit test):

epoch_no,train_loss,train_accuracy_global,train_f1_global,
0,0.445031365,[1.0, 0.0],[0.936170220375061, 0.0]

Expected behavior
This creates an issue when trying to parse the CSV, since the location of the actual results get moved. There are 2 potential solutions:

1. Either the class value needs to be appended in the metric, such as:

   epoch_no,train_loss,train_accuracy_global_0,train_accuracy_global_1,train_f1_global_0,train_f1_global_1
   0,0.445031365,1.0,0.0,0.936170220375061,0.0
   

2. Concatenate the metrics that give lists into a single string:

   epoch_no,train_loss,train_accuracy_globa,train_f1_global
   0,0.445031365,1.0_0.0,0.936170220375061_0.0
   

My expectation is that solution 2 is easier to implement and maintain, and 1 needs a wrappers per metric.

Screenshots
N.A.

GaNDLF Version

0.0.17-dev

Desktop (please complete the following information):
N.A.

Additional context
Related to #680

Expected behavior
A clear and concise description of what you expected to happen.

Screenshots
If applicable, add screenshots to help explain your problem.

GaNDLF Version

Version information of the GaNDLF package in the virtual environment.

Desktop (please complete the following information):

• OS: [e.g. Windows/Linux (include detailed distro information)/macOS]
• Version (including Build information, if any): [e.g. Fedora 22 or Windows 10.1803]

Additional context
Add any other context about the problem here.