[tree-ssa] Tail recursion optimisation

[Jan Hubicka]

> Hello,
> 
> this patch makes tail recursion to be eliminated already on trees, thus
> exposing more optimisation possibilities for tree-ssa optimizers.  Just
> very simple for now, I am going to improve it.  Bootstrapped on i686.

This is cool (one thing I wanted to implement for a while to familiarize
myself with tree-ssa branch).  Given that you can notice tail calls, it
would be nice to extend CALL_EXPR by extra flag specifying whether it is
tail or not.  We can use it later in expand_call to avoid
CALL_PLACEHOLDER hack...

Can tail recursion happen very early in the process? (ie before
tree_inline_function is done)  I would like to implement recursive
inlining and for that I must avoid tail recursive calls.

Honza