ASYNCIFY_ADVISE should output a tree

Currently if you use ASYNCIFY_ADVISE you'll get output that looks like this:

[asyncify] emglken_fill_stdin_buffer is an import that can change the state
[asyncify] __wrap_getc can change the state due to emglken_fill_stdin_buffer
[asyncify] data_raw_blockread_sub can change the state due to __wrap_getc
[asyncify] data_raw_blockread can change the state due to data_raw_blockread_sub
[asyncify] glk_select can change the state due to data_raw_blockread 
[asyncify] getglkchar can change the state due to glk_select
[asyncify] os_waitc can change the state due to getglkchar
[asyncify] os_getc can change the state due to getglkchar
[asyncify] bifcvtstr can change the state due to initial scan
[asyncify] biffopen can change the state due to initial scan
[asyncify] bifupr can change the state due to bifcvtstr
[asyncify] biflwr can change the state due to bifcvtstr

It would be more useable if it could be outputted as a tree, something like this:

[Asyncify] Functions that will be instrumented:

emglken_fill_stdin_buffer (EM_ASYNC_JS)
> __wrap_getc
  > data_raw_blockread_sub 
    > data_raw_blockread 
      > glk_select 
        > getglkchar 
          > os_getc (appears 2 times)
            > os_get_event
          > os_waitc 
      > os_getc (see above)
bifcvtstr (makes indirect call)
> bifupr 
> biflwr 
biffopen (makes indirect call)

So for each base function give the reason, then put all the callers in a nested tree below them. Some functions are in the tree multiple times - when that happens list them each time, but only display their callers once.

The main advantages would be it would be easier to get a handle on the call tree and which functions might be ASYNCIFY_REMOVE candidates, and that if a function is in the call tree multiple times that information is currently not output.