Optimise CBOR parsing + serialization through raw APIs and Codable

[Joannis]

Relies on #1

Makes an effort to:

• Make less copies to and from Data.
• Allow use and reuse of existing of [UInt8] values
• Encoding into an existing buffer, which can reduce necessary (re)allocations
• Optimise CBOR Integer parsing which was copying out the bytes into an array before
• Use lower defaults for maximum array/map/string/bytes sizes
• Replace blanket the indirect enum with an indirect case where relevant, as indirect makes the subject a reference counted value with heap storage