Testers, filters, and global functions should be traits

[SergioBenitez]

At present, all of these have signatures that require either static function (testers, filters), or static function/closures (global functions). Instead, all of these should use traits so that any structure with arbitrary state can be used as a test/filter/global function handler. Here's what I'm imagining:

trait Tester: 'static {
    fn(&self, _: Option<Value>, _: Vec<Value>) -> Result<bool>;
}

trait Filter: 'static {
    fn(&self, _: Value, _: HashMap<String, Value>) -> Result<Value>;
}


trait GlobalFn: Send + Sync + 'static {
    fn(&self, _: HashMap<String, Value>) -> Result<Value>;
}

For backwards compatibility and ease-of-use, you can provide a blanket impl for fn-like types. For instance, for Tester, we can do:

impl<F: 'static> Tester for F where F: Fn(Option<Value>, Vec<Value>) -> Result<bool> {
    fn(&self, a: Option<Value>, v: Vec<Value>) -> Result<bool> {
        self(a, v)
    }
}

[Keats]

Did it come up as something asked for Rocket? That seems ok to me.

[Keats]

@SergioBenitez Do you have a concrete example for the changelog from actual use? I'll probably implement it this week

[remexre]

What'd the status on this end up being? Did there turn out to be some hidden pitfall to using traits?

[Keats]

I don't think there are any pitfalls, just didn't have the time/forgot about it. I would take a PR for that though.