[PATCH] chg: make timeout adjustable

[Jun Wu]

Excerpts from Yuya Nishihara's message of 2016-06-14 23:20:49 +0900:
> Is it mainly for automated tests?

No. We got several user complaints about the timeout error and I found their
machines were overloaded. I plan to set CHGTIMEOUT=0 internally to avoid
future complaints. It makes sense since it's guaranteed that the hg
executable shipped is legit so I can count on "if hg serve runs, it must
create the socket file".

> I was slightly afraid of this, but it seems fine to expect the original
> timeoutsec is kept unmodified on failure.
> 
> http://stackoverflow.com/questions/25715204/ 

I used the pattern a lot before therefor didn't think twice about it.
Fortunately the standards are sane in this case.