Chapter 4: Scanning

[jeschkies]

This implements the token scanning.
I tried to stay as close to the original Java implementation as possible while still
being "rusty".

Differences to the Java implementation:

• There is no null check and a few if-else blocks became unwrap_or'('\0').
• Literals are not of type Object but use String and f64.
• is_digit and is_alpha are much simpler in Rust.
• The KEYWORDS map is not only static but create at compile time thanks
  to phf.

Possible improvements in Rust:

• One could use a PeekableIterator for Chars. However, this would diverged from
  the original scanner quite a bit.
• One could strike for zero copy for lexemes. To keep it simple I stuck with String.
• The strangest part is creating a substring with

  self.source .get(self.start..self.current).expect("Source token is empty.");
  

  I was afraid that changing it would make it harder to relate to the original chapter.
• The scanner could be less stateful and avoid self.current and self.start.