- Add caching for high performance Sets/Gets. The CPU usage and performance of GetNode/SetNode calls are significantly reduced after this change.

- Improve JSON marshalling performance using go-json. This helps squeeze out the last bit of performance particularly for embedded devices.

- Shortcut expensive heap allocations. The repetitive heap allocations may increase GC pressure and increase CPU usage when being called frequently.

For the node cache implementation, the applications that use the `ytypes` library maintain the cache(s) on their own. The applications can optionally pass in a pointer of the `node cache` and get a performance boost. Passing in a pointer of the `node cache` is relatively cheap and has a negligible performance impact. This implementation adds optional fields to the APIs and the node arguments and doesn't introduce breaking changes.

**Note**: unit tests need to be added for `caching enabled` method calls.

... and add initial `node_cache_test`.

…sues

- Fixed the node cache not handling the container type `unmarshalGeneric` correctly
- Added a temp node cache work around for memory address changes of slice heads when
  appending to a slice

Check nil for node cache stored `schema` and `parent`. An `Internal`
error code is returned if either `schema` or `parent` is nil.