deprecation warning P0, P1, P2

[Leopold Toetsch]

Here's again a summary of upcoming changes. I'll commit for now just the 
new "returncc" opcode. Other changes that start breaking things will 
follow in a few days.

Summary:
1) PIR code using PIRs HLL call syntax doesn't need changes.
   foo()         # call sub foo
   a = bar(x)    # call bar with argument, get return value
   (a,b,c) = baz(x,y)
   .return()     # return from a sub - maybe omitted if last
                 # statement before .end
   .return(foo, bar)     # return these values
   .yield(foo)   # yield value(s) from a coroutine
   self."method"(x, y)    # method call on object self
   $P0 = other_object     # and on another one
   (r, s) = $P0."other_meth"(a)
2) PASM code using pdd03 calling conventions
All explicit usage of P0, P1 related to subroutines as well as P2 in OO 
code is *not* safe and need changes.

2a) return from a sub
instead of "invoke P1"   use the new opcode "returncc"
2b) accessing the object "self" in methods:
    .include "interpinfo.pasm"
    interpinfo P2, .INTERPINFO_CURRENT_OBJECT
    # get object into given register
2c) access the current subroutine object
    .include "interpinfo.pasm"
    interpinfo P0, .INTERPINFO_CURRENT_SUB
2d) create a real continuation from the current continuation
    .include "interpinfo.pasm"
    interpinfo P1, .INTERPINFO_CURRENT_CONT
This continuation is already a full Continuation PMC usable for call/cc.
Reference:
docs/pdds/pdd03_calling_conventions.pod
I'll try to make these changes seamlessly, but I can't guarantee that 
there will be no intermittend failures in tests or other code, until the 
changes are done.

leo