S_newONCEOP: Rejig the op_next padop (Closes #18630)

[richardleach]

Two related changes pertaining to the padop executed when an OP_ONCE is encountered for the nth time.

• S_newONCEOP: correctly copy array/hash at subroutine exit
  Should a state array or hash be declared immediately prior to
  returning, the error Bizarre copy of HASH in subroutine exit error
  would be triggered. Now, the array or hash is returned without error.
  Closes Returning state hash-assign risks "Bizarre copy of HASH in subroutine exit" #18630

• S_newONCEOP/Perl_scalarvoid: skip over padops uselss in void context
  Rejigs how the nth-time padop is handled in void context to avoid the
  warning "Useless use of private variable in void context". The OPf_STATE
  flag is removed and the op_next chain now explicitly skips the padop.

---

• This set of changes requires a perldelta entry, and it is included.

[tonycoz]

An array in scalar context is normally the number of elements in the array, but that doesn't appear to be the case for the trailing state @foo:

$ ./perl -Ilib -E 'sub f { state @x = @_; } say f(3,2,1); say scalar f()'
321
1
$ ./perl -Ilib -E 'sub f { state @x = @_; } say scalar f(4,2,1); say scalar f()'
3
1

[richardleach]

Thanks, I'll look into it.

I see that in:

./perl -Ilib -E 'sub f { state @x = @_; @x; } say f(3,2,1); say scalar f()'
321
3
$ perl -E 'sub f { my @tto =(3,2,1) } say f(3,2,1); say scalar f()'
321
3
$ perl -E 'sub f { my @tto =(3,2,1); @tto } say f(3,2,1); say scalar f()'
321
3

Though not in:

$ ./perl -Ilib -E 'sub f { (3,2,1) } say f(3,2,1); say scalar f()'
321
1

[richardleach]

Looks like it comes down to clobbering (or not) OPf_WANT. I'll update the PR shortly.

[richardleach]

Ok, I think it has the correct behaviour now.