Change coffeescript flags so that code is evaluated, not just compiled

[jlburkhead]

Currently when using engine="coffee" the code is run with the -p flag which prints the compiled code instead of the output.

For example, the following .Rmd file

```{r engine="coffee"}
x = 42
console.log "The answer is ", x
```

compiles to the following markdown

```coffee
x = 42
console.log "The answer is ", x
```

```
## (function() {
##   var x;
##
##   x = 42;
##
##   console.log("The answer is ", x);
##
## }).call(this);
```

By removing the -p flag the code is evaluated and the resulting markdown is

```coffee
x = 42
console.log "The answer is ", x
```

```
## The answer is  42
```

which seems to be the desired output given my experience using knitr with R and python.

[ramnathv]

Interesting. My use case usually is to generate javascript that will dynamically render something on the page as opposed to using coffeescript as a static computational engine. I presume the easiest way to accommodate flags is to allow them to be passed on as chunk options.

[ghost]

By default it will end up in a code tag in the rendered html, so it won't actually run. Do you have an example of using it to dynamically render something? It still seems like -p should be removed from the default behavior for coffeescript to keep things consistent with the other engine options.

[ramnathv]

I always use it with comment = NA, results = 'asis'. A good example would be d3 code to create an interactive visualization. In this case you want the output to be javascript.

[yihui]

It makes more sense not to force -p by default, since it is possible to set engine.opts = '-p' if the user wants to print JavaScript. If I force -p, it is not possible for the user to cancel it, hence I have merged this PR. Thanks!