## Summary
This PR updates the string flags to include an `Invalid` variant for any
invalid string nodes as deemed by the parser. This is to avoid dropping
the nodes and instead just mark it as invalid. The nodes will be empty
for now but we can discuss on whether to keep the raw source text or
not. It's not really required because the range can be used to get the
same.
It also adds a new `handle_implicitly_concatenated_strings` method which
is similar to existing `concatenated_strings` function. The reason to
have a separate method is to avoid dropping all strings if there's an
error. The error being that it's concatenating bytes and non-bytes
literal. Now, we need to decide which strings to retain. Currently, I've
kept it simple to retain bytes literal _only_ if all of them are bytes
otherwise we'll have string / f-string with invalid nodes instead of
bytes literal.
This removes the need for having a `StringType::Invalid` variant.
## Test Plan
- [x] Existing test cases pass
- [x] No performance regression
1ec7259116