Make using property lists easier

[madsmtm]

It might make sense to add helper methods to AnyObject and CFType to allow more easily converting to common property list types (CF docs, Cocoa docs, core-foundation source).

Something like the following:

/// Property list helpers.
impl [AnyObject|CFType] {
    pub fn as_string(&self) -> Option<&[NS|CF]String> {
        self.downcast_ref()
    }

    pub fn as_number(&self) -> Option<&[NS|CF]Number> {
        self.downcast_ref()
    }

    pub fn as_data(&self) -> Option<&[NS|CF]Data> {
        self.downcast_ref()
    }

    pub fn as_array(&self) -> Option<&[NS|CF]Array> {
        self.downcast_ref()
    }

    pub fn as_dict(&self) -> Option<&[NS|CF]Dictionary> {
        self.downcast_ref()
    }

    pub fn as_date(&self) -> Option<&[NS|CF]Date> {
        self.downcast_ref()
    }

    // Only on CFType
    pub fn as_bool(&self) -> Option<&[NS|CF]Boolean> {
        self.downcast_ref()
    }
}

An alternative would be to wrap these in a newtype that provided much the same, but I suspect that'd be less ergonomic. Yet another alternative would be to eagerly convert to an enum, but that'd probably be less performant.

We should also make sure it's easy to use dictionary getter methods here.

[madsmtm]

Something to discuss would be whether to eagerly convert the NSNumber to a float and/or integer? Something like:

impl AnyObject {
    pub fn as_f64(&self) -> Option<f64> {
        let number = self.downcast_ref::<NSNumber>()?;
        if number.contains_encoding::<f64>() {
            Some(number.as_f64())
        } else {
            None
        }
    }

    // Same for other numbers.
}

But it's unclear whether the contains_encoding check would be desirable here or not?

[madsmtm]

A benefit of the newtype would be that it might be possible to make safe? Depending on how much CF-like frameworks verifies the types of objects before using them?

See also #562.