CSV: don't eagerly check next char after newline

[asterite]

Fixes #11172

When a \n char was detected by the lexer, it would check what the next char was. If it was \0 it means the end of the file/stream as opposed to a newline. However, doing this when streaming over an IO meant that the newline token wasn't returned until more content was available after the newline. This isn't ideal.

This PR changes that so that we only check the content after the newline if a next token is asked. This allows the parser to produce a row right after a newline comes.

This is technically a breaking change because something like "a\n" would previously produce Cell("a"), EOF but now it will produce Cell("a"), Newline, EOF. I think this breaking change is acceptable because:

• it's very unlikely to break client code (I doubt anyone is using the lexer instead of the high-level code)
• it can be considered a bugfix: now you can distinguish between "a\n" and "a", if that matters to you
• the parser will still produce the same output

[sdogruyol]

Thank you @asterite 🙏