[krperry]

When an unspaced string such as a URL is longer than a whole line,
BrailleBlaster cut it at the end of the line with nothing to mark the
division, sometimes inside a word ("p]sonal-bank" / "+4html"). UEB
requires the line continuation indicator, dot 5 ("), at the end of
each line that divides such a string.

Changes (utd)

• LineWrapper gains a divideLongWord hook. The default returns
  null and keeps the old cut.
• New UebLineWrapper divides at the last natural point that fits
  with room for the indicator: after _/ or a hyphen, or before a dot
  (4) or @a. Failing that, it takes the last place that keeps a
  prefix cell with its cell and a number sign with its digits. Spaced
  text, and Unicode text that isn't Braille cells, keep the old cut.
• PageBuilder uses UebLineWrapper for literary text only when the
  main table is UEB. EBAE, Spanish, Cherokee and the others are
  unchanged, and math keeps its own wrapper.
• A string that fits whole on the next line still moves there.

BrailleBlaster still moves a long URL to a new line before dividing
it, so its division points differ from Duxbury's, which starts the
URL on the current line.

Testing

• New UebLineWrapperTest: division at a slash and before a dot, not
  inside a word; prefixes and numbers kept whole; a string that fits
  on the next line moves without the indicator; a URL over three
  lines; non-UEB tables unchanged; spaced text not divided.
• FormattingTest.longWordLineWrapTest now expects the UEB division
  with the indicator.
• check_continuation.py: all 6 continuation statements pass. BR-102
  and BR-78 now show the indicator, and every page fits 40 cells by
  25 lines.
• Regression statements unchanged. BR-83_acbs differs from its old
  baseline only by the bullet removal from Fix List Paragraph bullets and lone list bullets in DOCX import #176.
• mvnw package with tests passes.