Investigate feasibility of UTF8StreamJsonParser without canonicalization

[carterkozak]

This is a more specific follow-up from one of the components of FasterXML/jackson-benchmarks#6

I'd like to determine whether this comment still holds true, initial benchmarking showed a fairly substantial improvement for small inputs, and the UTF8StreamJsonParser has optimizations outside of keys that I'd expect to tip the scales in its favor.

jackson-core/src/main/java/com/fasterxml/jackson/core/json/ByteSourceJsonBootstrapper.java

Lines 259 to 270 in 36cd882

	if (enc == JsonEncoding.UTF8) {
	/* and without canonicalization, byte-based approach is not performant; just use std UTF-8 reader
	* (which is ok for larger input; not so hot for smaller; but this is not a common case)
	*/
	if (JsonFactory.Feature.CANONICALIZE_FIELD_NAMES.enabledIn(factoryFeatures)) {
	ByteQuadsCanonicalizer can = rootByteSymbols.makeChild(factoryFeatures);
	return new UTF8StreamJsonParser(_context, parserFeatures, _in, codec, can,
	_inputBuffer, _inputPtr, _inputEnd, bytesProcessed, _bufferRecyclable);
	}
	}
	return new ReaderBasedJsonParser(_context, parserFeatures, constructReader(), codec,
	rootCharSymbols.makeChild(factoryFeatures));

[cowtowncoder]

+1 for this; hopefully for 2.16 timeframe (I saw PR).