#wishlist speed up Uniform Decomposition

[caranha]

There is just so much speed up that can be done, since it is O(N^3*nobj). But I believe we can still make it a bit faster. Here is the output from Rprof:

> Rprof(tmp<-tempfile())
> x <- decomposition_uniform(list(N=100,.nobj=4))
> Rprof()
> summaryRprof(tmp)$by.total
                               total.time total.pct self.time self.pct
"decomposition_uniform"            163.56     99.99      0.00     0.00
"apply"                            163.54     99.98    105.98    64.79
"which.min"                        163.54     99.98      0.00     0.00
"FUN"                              163.50     99.95     26.50    16.20
"cd2"                              163.44     99.91      0.04     0.02
"sapply"                           161.42     98.68      0.06     0.04
"lapply"                           160.18     97.92      1.42     0.87
"aperm"                              6.80      4.16      3.80     2.32
"matrix"                             5.64      3.45      4.38     2.68
"unlist"                             3.94      2.41      3.94     2.41
"aperm.default"              
(... skipped the rest ...)