Q1 Should we ordinal or nominal encoding if we have extremely negative, negative, neutral, positive, or extremely positive as labels
Q2 Which metrics will be more important with the labels mentioned above, Precision, Recall, or F1Score? It is multi-class classification problem
If you’re doing classification, the labels should be one-hot encoded.
F1 score seems to be a popular choice, since it combines the other two.
Hello @vivek16pawar !! Su here! 
Regarding question 1, I think One-Hot Encoding would be a good match. For the second question, I personally would utilise the F1 Score because you technically it balances, I would say brings Precision & Recall together in harmony, basically to optimise the balance of the two aspects.