Document use cases and make sure we have examples covering them all

use cases for a parser of serialized data

There are some extremes:

* full deserialization allocates a lot of memory, but then you are free to do whatever you want
* copying without parsing
* probabilistic filtering, for example using simd instructions to check certain strings appear in a JSON file before attempting to parse it and double check that it is actually there

We are optimizing for the non extreme cases, while also supporting full deserialization. These use cases want to limit allocations, but still be accurate (no false positives). This includes:
* filtering (validation)
* partial deserialization of specific fields (capturing)
* transcoding (this includes deserialization as it transcodes into a memory representation)
* partial copying (only copying specific fields or removing specific fields) (this falls under partial deserialization and transcoding into the original format)