Non-exclusive tokens lookup is wrong when there are paddings

[koheiw]

> toks <- tokens("a x c y d") %>% 
+   tokens_remove(c("x", "y"), padding = TRUE)
> toks
tokens from 1 document.
text1 :
[1] "a" ""  "c" ""  "d"

> tokens_lookup(toks, dictionary(list("aa" = "a", "bb" = "b")))
tokens from 1 document.
text1 :
[1] "aa"

> tokens_lookup(toks, dictionary(list("aa" = "a", "bb" = "b")), exclusive = FALSE)
tokens from 1 document.
text1 :
[1] "AA" "AA" "c"  "AA" "d"