fix(gnovm): use OpCall to run deferred statements

[thehowl]

This is an itch that started when noticing something reviewing #2145, and then continued when breaking down #1672, and here we are.

Essentially, currently on master using defer will not create a new *Frame. This results, in #2145, in the deferred function not showing up in the stacktrace. doOpReturnCallDefers (the function in charge of executing deferred functions) actually has its own copied implementation of OpCall.

I haven't been able to fully track down the implications of not creating a new frame (zrealm_std7_stdlibs.gno is an attempt at doing so, but is actually correct on master, so 🤷). But what I do know is that having two copied implementations leads to bug being fixed in one case, and not in another one: to prove this, see defer10.gno.

defer10's case is fixed on OpCall, as a previous commit moved it from using copy(b.Values, dfr.Args) to using b.Values[i] = pv.Copy(m.Alloc) over all the values instead. (ie. copying values by reference when calling a function).

So, by always using OpCall to do everything, we create the frame correctly, and use a unique implementation to actually call functions.

I also modified the logic to determine whether a recover() can correctly recover a panic or not. (cc @deelawn). Instead of keeping a Scope counter in the Machine, the Exception now contains a slice of the frame pointers at the time of creation of the exception. The frame will be considered as being run during a panicking sequence only if it is in this array. We use this both for recover(), and for doOpReturnCallDefers, too.

@deelawn I tried using your system, but I couldn't get it working on a case like recover12b, with the new frame added. And I'm still not 100% sure how the system you had worked. Do you think this is ok?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 58.74%. Comparing base (c82e646) to head (350a98e).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2597      +/-   ##
==========================================
- Coverage   60.21%   58.74%   -1.48%     
==========================================
  Files         562      560       -2     
  Lines       75031    74345     -686     
==========================================
- Hits        45179    43671    -1508     
- Misses      26476    27370     +894     
+ Partials     3376     3304      -72     

Flag	Coverage Δ
contribs/gnodev	61.40% <ø> (+0.81%)	⬆️
contribs/gnofaucet	14.46% <ø> (ø)
misc/genstd	80.54% <ø> (ø)
misc/logos	20.23% <ø> (ø)
tm2	61.97% <ø> (-0.10%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[thehowl]

Btw this is against the current direction of #2596, as it heavily relies on the "frame as a pointer" system. Maybe there's a way to do this without Frame being a pointer.

[thehowl]

Looks like there's some CI trouble, back to draft for now.