Made the changes in comment #1 and pushed here: https://hg.savannah.gnu.org/hgweb/octave/rev/2eabd2567846

(There are more style changes to make to the BCI branch over the next few months.)

After discussions in the discourse thread about aligning the stack frame objects of the bytecode interpreter with how the non-bytecode frames work, I've done some much needed refactoring of the bytecode_fcn_stack_frame class.

The bytecode_fcn_stack_frame class was quite entangled with back and forth handling of scripts, nested, anonymous function etc, so I tried to make it cleaner by splitting it up and removing redundant code.

Refactor dynamic stack frame for bytecode interpreter

Split up the bytecode_fcn_stack_frame into fcn, script and nested
to detangle unessecary messiness.

Also solve some bugs where handles to nested function did not look
for symbols in the correct frame and did not push the closure frame
properly.

* call-stack.cc: Remove redundant check for bytecode frames
                 Use new stack_frame::create_xxx fcns
* stack-frame.cc: Refactor bytecode_fcn_stack_frame
* stack-frame.h: New create_bytecode_nested ()
                 New create_bytecode_anon ()
                 New create_bytecode_script ()
* ov_base.h: New enum value OCT_NESTED_HANDLE
* ov-fcn-handle.h: New get_closure_frame ()
* ov-fcn-handle.cc
* ov-ref.cc: Use offset overload
* ov.h: More friends
* pt-bytecode-vm.cc: Handle bytecode calls to nested handles
* pt-butecode-walk.cc: Remove fake offset workaround
* bytecode_nested.m: Update tests