Auto merge of servo#189 - tromey:remove-confusing-dead-code, r=SimonSapin

Clarify some dead code

next_token has this code:

            if is_ident_start(tokenizer) { IDHash(consume_name(tokenizer)) }
            else if !tokenizer.is_eof() && match tokenizer.next_byte_unchecked() {
                b'a'...b'z' | b'A'...b'Z' | b'0'...b'9' | b'-' | b'_' => true,
                b'\\' => !tokenizer.has_newline_at(1),
                _ => !b.is_ascii(),
            } { Hash(consume_name(tokenizer)) }

I noticed that the `_` case incorrectly checks whether `b` is ASCII,
rather than the current byte.

However, on further inspection, it turns out that this can't be run,
because all such cases would have been filtered out by is_ident_start.

This patch clarifies this code by removing the dead cases and adding a
comment.

<!-- Reviewable:start -->
---
This change is [<img src="https://reviewable.io/review_button.svg" height="34" align="absmiddle" alt="Reviewable"/>](https://reviewable.io/reviews/servo/rust-cssparser/189)
<!-- Reviewable:end -->

Author: bors-servo

src/tokenizer.rs

@@ -493,9 +493,9 @@ fn next_token<'a>(tokenizer: &mut Tokenizer<'a>) -> Result<Token<'a>, ()> {
             tokenizer.advance(1);
             if is_ident_start(tokenizer) { IDHash(consume_name(tokenizer)) }
             else if !tokenizer.is_eof() && match tokenizer.next_byte_unchecked() {
-                b'a'...b'z' | b'A'...b'Z' | b'0'...b'9' | b'-' | b'_' => true,
-                b'\\' => !tokenizer.has_newline_at(1),
-                _ => !b.is_ascii(),
+                // Any other valid case here already resulted in IDHash.
+                b'0'...b'9' | b'-' => true,
+                _ => false,
             } { Hash(consume_name(tokenizer)) }
             else { Delim('#') }
         },