D3665: graph: improve graph output by using Unicode characters

[Yuya Nishihara]

>   This extension beautifies log -G output by using Unicode characters.
>   
>   A terminal with UTF-8 support and a monospace Unicode font are required.

First, I'm not sure if we'll want to maintain this extension in core since
unicode generally sucks.

That said, can you fix style nits detected by test-check-code.t and add some
tests so that anyone loving unicode outputs can start reviewing and queue this?

> +def extsetup(ui):
> +  def convertedges(line):
> +    def prettyedge(before, edge, after):
> +      if edge == '~':  return '╧'
> +      if edge == 'X':  return '╳'
> +      if edge == '/':  return '╱'
> +      if edge == '-':  return '─'
> +      if edge == '|':  return '│'
> +      if edge == ':':  return '┆'
> +      if edge == '\\': return '╲'

> +      if edge == '+':
> +        if before == ' ' and not after  == ' ': return '├'
> +        if after  == ' ' and not before == ' ': return '┤'
> +        return '┼'

> +    if node == 'o': return '◯'
> +    if node == '@': return '⌾'
> +    if node == '*': return '∗'
> +    if node == 'x': return '◌'
> +    if node == '_': return '╤'

These non-ASCII characters will have to be expressed in `\\xNN' form for
Python 3 compatibility. Any strings in Mercurial are bytes.

> +  def outputprettygraph(orig, ui, graph, *args, **kwargs):
> +    if not ui.plain('graph') and sys.stdout.encoding == "UTF-8":

s/sys.stdout.encoding/encoding.encoding/

The canonical encoding source is `encoding.encoding`. And we'll have to check
`encoding._wide` because most of the glyphs used here are "East Asian Ambiguous"
characters.

https://www.unicode.org/reports/tr11/tr11-33.html