Add new rule to keep whitespace between ascii tokens

This resolves most of the weird spacing around ASCII input. Note that it
has to happen at the end of the processing pipeline to work with rules
that make sure ascii-ish punctuation next to Japanese gets the right
results.

cutlet/cutlet.py

@@ -248,6 +248,13 @@ def romaji_tokens(self, words, capitalize=True, title=False):
                 out.append(tok)
                 continue
 
+            # preserve spaces between ascii tokens
+            if (word.surface.isascii() and
+                nw and nw.surface.isascii()):
+                use_space = bool(nw.white_space)
+                out.append(Token(word.surface, use_space))
+                continue
+
             out.append(tok)
 
             # no space sometimes