Handle special attributes

[lionel-]

Currently melt_dataframe() creates names attributes that aren't necessarily the same length as the output vector, which makes the print method crash.

# Can only create named column vectors with dplyr
df <- dplyr::data_frame(col1 = c(a = 1, b = 2), col2 = c(a = 10, b = 20))
df <- tidyr::gather(df, "col", "value")

str(df$value)
## [1] "numeric"
##  Named num [1:4] 1 2 10 20
##  - attr(*, "names")= chr [1:2] "a" "b"

print(df$value)
##  *** caught segfault ***
## address 0x400000ae, cause 'memory not mapped'
##
## Traceback:
##  1: print.default(df$value)
##  2: print(df$value)
##
## Possible actions:
## 1: abort (with core dump, if enabled)
## 2: normal R exit
## 3: exit R without saving workspace
## 4: exit R saving workspace
## Selection:

This fix concatenates the names of the melted vectors so that the output is appropriately named.

It also takes care of dim and dimnames because melt_dataframe() uses SET_ATTRIB wihch bypasses the sanity checks that are normally triggered when modifying those attributes from R, and thus also break various consistency assumptions. These attributes are now simply discarded.