Use this in the VIM mode to make the cursor wide in escape mode.

This makes it easier for alternative keymaps to hook into it.

$ now goes not till the end of line, but one symbol before it (so e.g.
$i works as in vim now)

define per-character key-bindings in a function, instead of a loop
(where local variable "m" was bound to the last value in loop - see
line 203 in original source - so the prefix-d' and prefix-y'
key-bindings were useless)

do not raise js error when mark is not set (for prefix-d' and
prefix-y' maps)

Added a list of implemented and major not implemented key-bindings at
the top

Added commands: de, dd, P, f, t, df, dt, F, T, dF, dT, cf, ct, cF, cT,
gg, s, ce, cb, r, D, cc

I hope to merge xmlpure into xml real soon now.

Closes #505

hideEnd param added to brace & tag finder funcs to allow end or not on folding
Finding by brace and tag now also behave in same way

Make character bindings work for maps that define a nofallthrough.

| is not fully correct (works on literal characters, not screen characters, so tab will provide different semantics).
^ is fully correct, I believe.
I has had its semantics fixed (to go to the correct place in the line).

This hack is both heroic and completely terrible: it adds
regexps for the various browsers that are used to detect situations
where line-wrapping is influenced by span boundaries. It then uses
HTML as similar as possible to the actual display HTML when measuring
character offsets, and inserts <wbr> elements to compensate for
the differences that the introduction of the measuring span might
introduce.

It seems to work, better than the old situation at least, but the
assumption that browsers use the same line-breaking algorithm across
versions and platforms is bound to bite us.

(Feature-detecting the equivalent of these regexps is possible, but
expensive -- takes about 200 milliseconds on Chrome, more on other
browsers.)

Closes #502

Closes #500

1) "colgroup" is not to be closed by another "colgroup" (there are other conditions in spec).
2) recursive self closing is allowed; example:
   <optgroup>
    <option>
   <optgroup>
3) selfclosers should also grab context; example:

   <optgroup>
    <option>
   <optgroup />

(Though it looks rather dead.)

It's broken again, and no one has complained about it. I guess
demand for IE6 is finally marginal.

Closes #494