Adding ggc_free to htab_create_ggc macro?

[Daniel Berlin]

Now that we have a ggc_free, should we use it in our htab_create_ggc 
macro as the htab_free.
I'm not talking about the  htab_delete function, I mean the last 
argument, which controls what the table itself is allocated/freed what, 
not the entries.
That would let the htab free the old ggc_alloc'd memory immediately 
when it expands the table.
We use htab_create_ggc in quite a few places, so it seems like it might 
be a win.
Has anyone benchmarked this change, and found it to be a loss, or was 
this just not explored yet?