Improve JSON parser demo

[oskarek]

There are a couple issues in the JSON parser demo (recently discussed in this pointfree Slack thread):

• Parsing characters built up of more than one utf-8 code point, such as 再, fails.
• The literal unicode code point format is faulty, lacking a literal u character before the numbers.

This PR fixes those issues, as well as extends the JSON parser to fully conform to the json standard. This means adding support for handling UTF-16 surrogate pairs, and exactly match the JSON specification's number format, instead of relying on Double.parser() which isn't exactly the same.

[stephencelis]

Thanks for the improvements here! Just one small thing to address before merging, I think.

[stephencelis]

Can you explain this change? I was under the impression that JSON/JavaScript only has the concept of a "number" and doesn't distinguish integer and float values.

[oskarek]

Hmm yes, true. JSON only has the concept of a number, which is a sequence of digits, optionally followed by a decimal point and a fractional part. Now, if we model this as just a Double in Swift, we lose information about the number. 1 and 1.0, for example, becomes indistinguishable. By modeling numbers that does contain a fractional part as a Double, and numbers that doesn't as an Int, we at least keep that information, which I think is often desirable in Swift.

To not lose any information we could model the numbers as a String (case number(String)), but that seems less convenient in practice. I think just separating numbers with and without any fractional part strikes a good balance of what information to keep about the numbers. This is at least how I'm thinking about this :)