v2.2.0

Under data/

• Updates the following languages, re-scraped with the corrected dialect
  selector (#593):
  • Bengali (ben)
  • Welsh (cym)
  • English (eng)
  • Armenian (hye)
  • Latin (lat)
  • Portuguese (por)
  • Spanish (spa)
  • Thai (tha)
  • Vietnamese (vie)

Under src/ and elsewhere

• Fixes the dialect XPath selector so that pronunciations nested under an
  accent header (as many common words are on Wiktionary) are matched by the
  accent label on their enclosing <li>, and refines the "no accent"
  fallback so that a general outer <li> no longer scoops the general
  pronunciations of its accent-specific sub-variants. (#593)
• Changes the default and English IPA extractors to select direct-child
  IPA spans instead of any descendant, so that each accent variant is
  extracted on its own line rather than the outer <li> also picking up the
  IPA of nested variants. (#593)
• Updates the dialect XPath selector for Thai,
  for the new Wiktionary's table layout. (#593)
• Adds verify_rescrape.py, a script for verifying a re-scraped TSV doesn't
  have major regression issues. (#593)