v0.7.10

If you validate EPUB 2 books, upgrade. Seven false positives fixed — cases
where epubveri rejected markup that epubcheck accepts. All were EPUB 3 rules
leaking into the EPUB 2 branch, found by auditing our rules against
epubcheck's own schema/20 instead of waiting for reports. No verdict changes
on the epubcheck test corpus.

Fixed

• <a name="…"> no longer draws RSC-005. The classic pre-id anchor
  form, so internal links in a large share of older EPUB 2 books were being
  rejected. OPS 2.0.1 assembles a's attribute list from four modules and
  epubveri carried only one of them; target, charset, rel, rev,
  shape, coords and usemap were missing for the same reason, and img
  was missing longdesc, name, shape and coords.
• An empty <title> is no longer an error on EPUB 2. XHTML 1.1 types
  <title> as text, which permits empty content; epubcheck asserts non-empty
  only for EPUB 3.
• lang and xml:lang may differ on EPUB 2. XHTML 1.1 declares them as
  independent attributes with nothing tying their values together.
• A nested <dfn> is no longer an error on EPUB 2 — that rule is EPUB 3
  only.
• The legacy <tours> package child is accepted, as opf20.rng allows.

Added

• Six NCX checks, completing epubcheck's ncx.sch: the playOrder
  sequence must start at 1 and have no gaps, elements naming the same target
  must share a playOrder, a pageTarget's value+type combination must
  be unique, and sibling navLabel/navInfo elements must not repeat an
  xml:lang. All RSC-005, and all reported per offending element.
• scripts/scan-shelf.sh — runs epubveri over a directory of real EPUBs and
  summarises verdicts, error-level findings, and the number of distinct
  messages per ID. The corpus cannot find false positives; real books can.

Full notes: CHANGELOG.md. Also on crates.io.