Initial checkin of i18n models for SmartSelection and Sharing.

Test: Models run in eval on the Pygmalion dataset:
/tmp/i18n/eval.de.txt selection_accuracy,0.566706
/tmp/i18n/eval.es.txt selection_accuracy,0.5820337
/tmp/i18n/eval.fr.txt selection_accuracy,0.595679
/tmp/i18n/eval.it.txt selection_accuracy,0.66727275
/tmp/i18n/eval.ja.txt selection_accuracy,0.48282966
/tmp/i18n/eval.ko.txt selection_accuracy,0.66267765
/tmp/i18n/eval.nl.txt selection_accuracy,0.6062462
/tmp/i18n/eval.pl.txt selection_accuracy,0.66325915
/tmp/i18n/eval.pt-PT.txt selection_accuracy,0.629105
/tmp/i18n/eval.ru.txt selection_accuracy,0.58198196
/tmp/i18n/eval.th.txt selection_accuracy,0.4747754
/tmp/i18n/eval.tr.txt selection_accuracy,0.6404066
/tmp/i18n/eval.zh-Hant.txt selection_accuracy,0.59681696
/tmp/i18n/eval.zh.txt selection_accuracy,0.58001304

Change-Id: Ie1a9579f41ba65d554d57b33c52e300195e1b165