  * Implemented `safe` option.  This suppresses output of raw
    HTML and potentially unsafe URLs.
  * Documented `smart` option.
  * Allow non-initial hyphens in html tag names.
    This allows for custom tags, which in HTML5 MUST contain
    a hyphen.  See jgm/CommonMark#239.
  * Revised HTML block parsing to conform to new spec.
  * Improved `smart_punct.txt` tests, and added some commentary
    (a proto-spec).
  * Improved `smart` handling of dashes.  We now process
    whole strings of hyphens in such a way as to ensure that
    we don't get a mix of hyphens and em and en dashes (#56,
    Brandon Frohs).
  * Dingus: Removed duplicated CSS line (Aurelio Jargas).
  * Dingus:  Make permalink record whether 'smart' was selected (#55).
  * Dingus:  Removed Makefile dependency on `html5-entities.js`.
  * Reset options before calling `inlineParser`.
    Otherwise, if the main parser's options have changed since
    it was instantiated, the new options won't be visible to
    the `inlineParser`.
  * Properly split on whitespace in HTML renderer (#54).
  * Fixed bench for new version of showdown (soomtong).
  * `processEmphasis`: renamed `potential_openers` -> `openers_bottom`,
    better logic for setting lower bound in `openers_bottom`.
  * Fixed emphasis/link parsing bug (#50).
  * Reset `this.column` on new parse.
  * Fixed test runner to handle visible tabs in spec HTML.
  * Improved detection of column with tabs in input.  Added
    `advanceNextNonspace` and `advanceOffset` functions.
  * Removed `detabLine` and tabs-to-spaces conversion.
  * Added `column` and `nextNonspaceColumn` to parser.
    Adjust these in `findNextNonspace`.
  * Added note on README about how to fetch dependencies.
  * Fix link reference definition edge case (Benjamin Dumke-
    von der Ehe).  If the reference seems to have a valid title
    that does not go until the end of the line, check if the
    reference becomes valid when discarding the title.
  * Fixed link label recognizer.  Allow a backslash-escaped
    backslash (#38).
  * Use `mdurl` instead of copied `encode.js`, `decode.js`.
  * Use `entities` from npm instead of `html5-entities.js`.
  * Rewrote `findNextNonspace` for speed.
  * Remove delimiters as soon as we see that they don't match
    anything (#43).  This fixes a performance bug for things like
    `"a_ " * 20000`.
  * Improved efficiency of `processEmphasis` (#43).  Don't look
    for openers again when we've already searched for openers
    of a certain type.
  * Allow literal (non-escaping) backslashes in link destinations
    (Benjamin Dumke-von der Ehe).
  * Added several pathological tests (#43).
  * Don't consider images to be self-closing in the XML writer
    (Benjamin Dumke-von der Ehe).  Image nodes in a CommonMark AST
    are containers, and as such the XML writer will always output
    a closing tag.
