This way we don't need to reimplement PartialEq for Element. It's also
way easier to get an attribute by name as we don't need to iterate over
every attribute to see if it exists.
The only side effect is that now, in the Debug output, attributes are
automatically sorted by names instead of being sorted by insertion
order.
Fixes#4
The order of attributes in an `Element` doesn't matter anymore.
`<elem a="b" c="d" />` and `<elem c="d" a="b" />` are now correctly
considered equal.
For that I had to derive `PartialOrd` and `Ord` for `Attribute`.
This allows us to sort cloned vectors of `Attribute` in the `PartialEq`
implementation and compare them instead of the struct `attributes`.
Fixes#3
Instead of adding the local_name of an attribute, if a prefix exists,
add prefix:local_name to allow users to retrieve it via the namespaced
key name.
For example, with this XML:
```
<?xml version="1.0" encoding="utf-8"?>
<root xml:lang="en" >
</root>
```
`root.attr("xml:lang").unwrap()` will now correctly return "en".
`root.attr("lang")` will not retrieve "xml:lang" value anymore.
This is a breaking change.
Fixes#2
