R-alpha: generic write()

 Attached is a first shot at a generic write with my `old' write.table as
 write.data.frame, suitably modified.  There is also a stupid generic
 latex function, which is just there because it illustrates that the
 extra arguments sep and eol are useful.

 In playing with it, I noticed a few `problems'.

 * The documentation for factor() says,

	If exclude is set to a zero length vector, then any NA values in
	x are used for form a new level for the factor.  This means that
	there will be no NA values in the result.

 Perhaps I don't really understand this, but

 R> x <- factor(c(1:3, NA), exclude=numeric(0))
 R> x
 [1]  1  2  3 NA
 R> levels(x)
 [1] "1" "2" "3"

 * coerce.c has two error messages of the form

	error("use \"factor\", \"ordered\", \"cut\" or \"code\" to create factors\n");

 but there is no function `code'.

 * cat() is a bit inconsistent:

 R> cat(x, sep = "&")
 1&2&3&4R> cat(x, sep = "&\n")
 1&
 2&
 3&
 4
 R>

 Actually, a final newline is added whenever sep contains a newline.

 >From src/main/builtin.c:

        nlsep = 0;
        for (i = 0; i < LENGTH(sepr); i++)
                if (strstr(CHAR(STRING(sepr)[i]), "\n")) nlsep = 1;

	...

        if ((pwidth != INT_MAX) || nlsep)
                Rprintf("\n");

 I am not sure this is the right thing to do.

 (In particular, it makes it rather messy to incrementally build data
 files with `explicit' end-of-line strings (such as "\\") by appending
 via cat().)