RFC: add <- as an assignment-like operator

[JeffBezanson]

We currently don't parse this at all. This might be useful to define on objects that support some kind of insertion or assignment operation. One motivation is that I've seen <= used for that purpose.

This change parses <- with assignment-level precedence but as an ordinary function call.

[martinholters]

We currently don't parse this at all.

Errr, well, before this, we had

julia> 0<-1
false

while now we have

julia> 0<-1

ERROR: UndefVarError: <- not defined

Is it worth reverting this and instead deprecating a<-b in favor of (the more readable, anyway) a < -b?

[JeffBezanson]

Oops, you're quite right. That was a mistake. I knew there had to be some reason we didn't have that operator...

I'll revert this, then we can discuss whether to deprecate. Another option is to use <--, which does give a parse error in 0.5.

[tkelman]

We need to have a test for the case that this broke.