Qhanzi upgrade and pen input testing site

[Ben Bullock]

I've done yet another upgrade on qhanzi.com consisting of fixes for some bugs and accuracy improvements. The current accuracy level looks like this:

Total 54456 tests, found: 53500 (98.244%), average position: 1.556.

Development is proceeding much more quickly due to adoption of a new testing regimen, codename "harue", which uses the Go language's reflect package to make testing of various parameters go much more quickly.

A testing site for pen inputs is available at

https://www.lemoda.net/tools/shape-snatcher/

It should look like this:

It works on touch and mouse inputs but since I don't have a pen input I am awaiting user feedback.

The silly name is a play on "shapecatcher.com". I'm planning to use the "mariko" algorithm mentioned previously to implement some kind of shape matching for general shapes beyond kanji.