python3: add raw prefix in cases harder to analyze at the token level The `byteify-strings.py` script would be a lot more complicated if it had to do backtracking and other more advanced static analysis to figure our those cases, so we have to add the raw prefix to those cases manually.

python3: add byte prefix for objects that look like kwargs but aren't The `byteify-strings.py` script has no way of knowing that those aren't acutally kwargs since it works purely at the tokenization level, so we have to add the byte prefix to their keys manually.