Great overhead for setTimeLimit?

Hi all,

From the document of 'setTimeLimit', it states "Setting any limit has

a small overhead ? well under 1% on the systems measured.", but
something is wrong with my benchmark code, enabling the time limit
makes my benchmark 1x slower than the benchmark without the limit.
Below is an example

```
benchFunc <- function(x, data) {
    value <- 0
    for(i in 1:5000){
        for(j in seq_along(data))
            value <- value + data[j]
    }
    value
}
data <- sample(1:10, 10)

setTimeLimit(Inf, Inf, FALSE)
system.time(lapply(1:5000, benchFunc, data = data))

setTimeLimit(999, 999, FALSE)
system.time(lapply(1:5000, benchFunc, data = data))
```

Here are the test results

setTimeLimit(Inf, Inf, FALSE)
system.time(lapply(1:5000, benchFunc, data = data))

user  system elapsed
 10.809   0.006  10.812

setTimeLimit(999, 999, FALSE)
system.time(lapply(1:5000, benchFunc, data = data))

user  system elapsed
 13.634   6.478  20.106


As a side note, it looks like the GC consumes the most CPU time. The
GC costs 10 secs without the time limit, but 19 secs with the limit.
Any thoughts?

Best,
Jiefei