Add "chunkwrap" metatag

[towerofnix]

Development:

• Towards Investigate wrapping solutions for extremely long "try not to break" elements #341.

Notes:

• Chunkwrap is magic!
• Chunkwrap is very pure and safe magic. Chunkwrap would never hurt you.
• Chunkwrap may not play nice with alternate localization, and this is left as a concern to address later.
• Chunkwrap takes inline-block wrapping responsibilities away from the parent and moves it into each of a sequence of children which chunkwrap creates, by splitting after each instance of a specified marker (for example , as in an accent). The parent gets no special wrapping behavior, at each child is treated as its own wrappable unit.
• Chunkwrap is currently used to make word wrapping in the additional names box nice.

Technical details:

• Apart from html.tag('chunkwrap', {split: ','}), this PR also introduces a new utility html.smush(). Smushing is a simple recursive operation which provides a nice, even, not-bumpy layer of top-level texts and tags. It takes hierarchies of content-only tags nested directly under other content-only tags, such as produced by language.formatString or any call to html.tags([...]), and evens them out into a single continuous layer. Any adjacent bits of text, including opposite boundaries of html.tags([]) containing text as well as tags, are combined into single strings, so they can be conveniently operated on.
  • Smushing always results in a hierarchy that stringifies to the same output as if the input hadn't been smushed.
  • Smush will traverse tags which are contentOnly, but never tags which are not. Smush doesn't care at all about what's going on inside a normal element; it only wants to smooth a run of apparently top-level tags and apparently top-level texts into one layer of actually top-level tags and texts.
  • Smush will resolve templates, but only at the top level. This is so that it can actually see everything that's going on at the top-level. If a template returns html.tags([]) which directly nests more templates, those will also be resolved, but only until everything at the top layer is a sequence of tags and texts.
  • Smushing never results in two strings next to each other, but it can result in two tags next to each other, if there wasn't any text between them.
  • Smushing respects the smushee's [html.joinChildren] attribute at each layer. This means it applies the appropriate joiner wherever stringifying would; each parent applies its joiner to between own children, not between the children that each recursively smushed child will actually be represented as.
• Smushing exists to address the "transparent" layers which we've made use of in these prior pull requests:
  • Dodge nasty HTML in data content more aggressively #269
  • Refactor string formatting across HTML and language, add "blockwrap" metatag #348
• The chunkwrap metatag smushes its own contents (i.e itself) so that it has direct access to the text components which divide its non-text chlidren. Chunkwrap never splits inside elements (nor inside attribute values!!), so it needs a clean hierarchy of just the top-level texts. Those are safe to split inside.