adds better support for dplyr::mutate + tests

[adgaudio]

This PR wraps dplyr::mutate the same way we do for filter and arrange so it's more usable in tidyjson expressions.

@jeremystan

Solves this bug:

>      key_employees <- companies %>%
+        as.tbl_json %>%
+        spread_values(
+          name = jstring("name")
+        ) %>%
+        mutate(
+          company.sort_order = rank(name)
+        ) %>%
+        enter_object("relationships") %>%
+        gather_array("relationship.index") %>%
+        spread_values(
+          is.past = jlogical("is_past"),
+          name = jstring("person", "permalink"),
+          title = jstring("title")
+        )
Error: is.tbl_json(x = x) is not TRUE

[jeremystan]

There is a documentation change to allowed types that may be because you ran document() using older code that hadn't been run with document? If so, that may be acceptable (though we should verify the documentation is correct).

[adgaudio]

Also, what do you think about those factor(...) calls in the test. Are we okay with them in there? (They are necessary to make tests pass).

[jeremystan]

Rather than having to call factor() above, consider either using data_frame (may create a different object than data.frame? not sure), or instead use data.frame(..., stringsAsFactors = FALSE).

[adgaudio]

@jeremystan

[jeremystan]

+1