You're going to want to install the Perl API.

Then run this script:

#usr!/bin/perl
use warnings;
use strict;

use Bio::EnsEMBL::Registry;

my $registry = 'Bio::EnsEMBL::Registry';

$registry->load_registry_from_db (
    -host => 'ensembldb.ensembl.org',
    -user => 'anonymous');

open (IN, "<inputfilename.txt");
open (OUT, ">protein_coding.txt");

my $ga = $registry->get_adaptor( "human", "core", "gene");

while (<IN>) {
    my @fields = split / /, $_;
    my $gene = $ga->fetch_by_stable_id($fields[0]);
    if ($gene->biotype eq 'protein_coding' or 'IG_V_gene' or 'IG_D_gene' or 'TR_C_gene' or 'TR_J_gene' or 'TR_V_gene' or 'IG_J_gene' or 'IG_C_gene' or 'TR_D_gene'){
        print OUT $_;
    }
}

You can edit the file names to match your files.

Don't use BioMart to try to extract whole genome data. It can't handle that amount of data and breaks down partway through your query, only giving you a partial dataset, but not giving you any warning that it's done it. What format do you have your output in, and what format are you looking to get?

If you know R, and your GTF file used to generate cufflinks output is based on ensembl ids, I don't think this is a difficult task for biomaRt at all. If I understand correctly, you just want to know which ids are protein coding, and then grab those results from your result files. Here is an example of doing that, assuming you can point to a directory containing directories of cufflinks results:

# biomaRt
library(biomaRt)
bm <- useMart("ensembl")
bm <- useDataset("hsapiens_gene_ensembl", mart=useMart("ensembl"))

# grab some annotation
anno <- getBM(attributes=c("ensembl_gene_id", "external_gene_id", "transcript_biotype", "description"), mart=bm)

coding <- c("protein_coding","IG_V_gene","IG_D_gene"
            ,"TR_C_gene","TR_J_gene","TR_V_gene"
            ,"IG_J_gene","IG_C_gene","TR_D_gene")

# narrow it down
anno.coding <- anno[anno$transcript_biotype %in% coding,]

# get a list of cufflinks directories
cufflinksDirs <- dir()

# create something to hold the data
fpkm <- data.frame(matrix(NA, nrow=nrow(anno.coding), ncol=length(cufflinksDirs)))
rownames(fpkm) <- anno.coding$ensembl_gene_id

for( i in 1:nrow(cufflinksDirs) ){ 
  fname <- paste(cufflinksDirs[i], "/genes.fpkm_tracking",sep="")
  # tell me
  cat(fname, "\n")
  # read in the data
  x <- read.table(file=fname, sep="\t", header=T, as.is=T)
  # match data ids to master table
  iv <- match(x[,1], rownames(fpkm))
  fpkm[iv,i] <- x[,"FPKM"]
}

colnames(fpkm) <- cufflinksDirs
write.table(fpkm, file="fpkm.txt", sep="\t", col.names=NA)