Automatic cost center insertion on overloaded bindings and overloaded call sites

Motivation

It should be possible to automatically add cost centers to overloaded bindings and overloaded call sites, after optimization. This would enable easier detection of overloaded calls that "survive" optimization. It could also help users determine which overloaded calls are most expensive, obviating candidates for specialization.

Proposal

Add two new cost center insertion methods enabled via -fprof-late-overloaded and -fprof-late-overloaded-calls. The former adding cost centers to top-level overloaded bindings, the latter adding cost centers to call sites involving dictionary arguments.