track JsonFactory in CharsToNameCanonicalizer

[pjfanning]

• not yet making a change for ByteQuadsCanonicalizer
• based on validate the length of names #1078 (comment)
• I don't really understand the subtleties in CharsToNameCanonicalizer. In validate the length of names #1078, I am already validating the length before I call findSymbol on the CharsToNameCanonicalizer so I don't know what benefit there is to also checking in CharsToNameCanonicalizer. The changes in ReaderBasedJsonParser in 1078 are incomplete because waiting until the CharsToNameCanonicalizer findSymbol call is very late. We may already have wasted a lot of processing time by that stage.

[cowtowncoder]

Point is not to have check in both places, but only in CharsToNameCanonicalizer. For performance reasons check for max length really should not/can not be in tight loop for decoding, unfortunately; things are bit easier for byte-backed case.

But FWTW I am mostly/only worried about byte-based side, decoding of quad-based input into String: not for collecting quads themselves (or char buffer). These can be detected right before processing collected raw input.
Some users might prefer more exact failure for exact limit, but that is very difficult to implement for byte-based input (at least wrt limit being defined as chars, if that was supported).

[cowtowncoder]

@pjfanning I implemented this slightly differently, see #1088. Closing this PR as not needed any more.