Rewrite rbx_xml to use quick_xml instead of xml-rs

[Dekkonot]

rbx-xml is slow and has trouble processing unicode, among other issues. This can and is solved by simply swapping to a new XML parser. Given the speed, I've decided to go with quick-xml.

Unfortunately, the current parser (xml-rs) is intertwined rather thoroughly with rbx-xml's implementation and swapping it out for quick-xml would be impossible without refactoring large segments of it. Given there's a few other design choices that I wanted to change, I decided to bite the bullet and go for a full rewrite. Initial testing is good and suggests we're looking at 15x the parsing speed in some cases. I have tried to prioritize not copying data when possible, so overall memory should also be low.

Update: This is mostly done now, I believe. Many many tests need to be written but the bulk of the code will not change. With that in mind, here are my thoughts beyond "reviewing this is a nightmare":

1. This rewrite fundamentally changes how the crate is laid out. Specifically, it's split serializer and deserializer into dedicated modules and the code for (de)serializing data types now lives under those modules. This differs dramatically from the current system we use that implements XmlType for each type and both deserialization and serialization in one place. In my opinion, this is easier to reason with but it is significantly different and may require us to point contributors in the right direction if they're used to the old system. Also, it means we have the potential for element name mismatches between serializing and deserializing. Testing will absolutely catch that, but it's something we didn't have to deal with before.

2. The deserializer now requires the referent attribute to be set on Item elements. I was very surprised to learn it was not before, but as a result this is also a breaking change. I don't think we should care though.

---

For posterity, here's a checklist of what needs to be done:

• Visit properties using the ReflectionDatabase
• Port data types
• Redo the serializer
• Run every test like a million times to ensure I didn't break anything

[Dekkonot]

Rather than work out what work I'm missing since I started this rewrite, I'm going to close this and instead focus on a less invasive rewrite. I think it will result in a much faster turn around for the port, and in turn it should end up being easier to review. Sad to see my work go, but it will not die here.