slide-maker 6.0.1 — a package with no pictures, and text measured as it renders

A package with no pictures works, every direction preview draws, and text is measured as LibreOffice sets it.
SkillHub accepts no image files; without the bundled pictures the direction preview, the runnable examples and the
contract check now still work. Twenty-two direction previews that raised in 6.0.0 draw again, and a round of layout
fixes measures figures, mixed Chinese and Latin lines, line ends and baselines the way the render sets them.

• A package with no pictures. Without assets/, the direction preview takes the library's copy of each sample
  from slides.addsum.top, laid over the palette cover (which shows if the picture cannot load);
  vl.sample_picture() returns the bundled picture or draws a plain stand-in for the examples, the sample decks and
  the contract's pages; the contract skips the bundled-sample check. With assets, nothing changes.
• Direction previews. Eleven languages (herbarium, airmail, vinyl, receipt, comic, bricks,
  circuit, topo, periodic, stainedglass, calendar) declared a cover the preview does not draw, and comic
  a skeleton it does not draw, so choosing them in the direction gate raised (22 of 60 language x ground previews).
  Their manifests now name covers the preview draws, and the contract refuses one it cannot. The style-sample label
  is dark on its white strip, readable on dark grounds.
• Measured as LibreOffice sets it.
  • Figures of an old-style face (Didot, Hoefler Text, Georgia) are measured in the lining face they are drawn in:
    "1:250" and "214" no longer break as "1:25 / 0" and "21 / 4".
  • CJK line ends follow the render: one closing mark hangs, an opening quote or bracket never ends a line, and a
    break moves back to keep a Latin word whole (a full-width comma after "WikiGPT-5" takes the whole word to the
    next line, not its last letter), in the layout and in the delivery lint.
  • In mixed CJK and Latin lines a space takes the script before it and the autospace between scripts is counted;
    bold and italic of .ttc families are measured in their own faces; a one-word field keeps the same headroom as
    the rest; credit chips hold their CJK lines.
  • vl_native2.stack() centres and bottoms the plan it made, with no false refusals, and first_baseline()
    follows the render's line model (pixel's coin stands on the digits' baseline again).
• Delivery lint. A card is a body that a drawing bounds (open traces, outline sets and halftone dots are no
  longer cards); vertical text is read as columns; a page's own frame band is not a column rule.
• README. The visual-language section sits under the template gallery: twelve looks, three to a row, each
  opening its language in the library, now a section of slides.addsum.top (#languages, /cn/#languages);
  /languages/ redirects there.

Full changelog: https://github.com/addsumtech/slides_maker/blob/main/CHANGELOG.md