[PPC] implement enter_interpreter_frame

[vchuravy]

This fixes parts of #26250, by implementing enter_interpreter_frame on PPC.

During the implementation I followed the PPC ELFv2 ABI as documented in 1.
I am not a 100% sure about the cfi directives, but GDB can now construct a proper
backtrace from enter_interpreter_frame.

This still doesn't pass the test for #23971

julia> let 
              for i = 1:1
                   global bt23971 = backtrace()
              end
         end

On PPC64:

julia> st = stacktrace(bt23971)
7-element Array{Base.StackTraces.StackFrame,1}:
 backtrace() at stacktraces.jl:189                                       
 top-level scope at REPL[1]:3                                            
 eval at boot.jl:317 [inlined]                                           
 eval(::Module, ::Expr) at REPL.jl:5                                     
 eval_user_input(::Any, ::REPL.REPLBackend) at REPL.jl:85                
 macro expansion at REPL.jl:116 [inlined]                                
 (::getfield(REPL, Symbol("##28#29")){REPL.REPLBackend})() at event.jl:92

On x86-64:

julia> st = stacktrace(bt23971)
6-element Array{Base.StackTraces.StackFrame,1}:
 top-level scope at REPL[1]:3                                            
 eval at boot.jl:309 [inlined]                                           
 eval(::Module, ::Expr) at REPL.jl:5                                     
 eval_user_input(::Any, ::REPL.REPLBackend) at REPL.jl:86                
 macro expansion at REPL.jl:117 [inlined]                                
 (::getfield(REPL, Symbol("##28#29")){REPL.REPLBackend})() at event.jl:92

Note the spurious backtrace on PPC. If anybody has an idea what part I am missing that would be appreciated!

[vchuravy]

Ok I think this implementation is correct and what I am seeing with backtrace is not related.