Restore the ABI function PyGen_New(). Previously Stackless named this function PyGenerator_New() and
used a macro the redefine PyGen_New as PyGenerator_New. Additionally I removed Stackless specific code
from PyGen_New(). Now Stackless uses a predefined frame of an exhausted generator to create new generators
during unpickling.

https://bitbucket.org/stackless-dev/stackless/issues/111
(grafted from 41e18aa4719f2a48bef350af817ad5314b9a0ae4)

…tacklesseval.c

The generator functions in stacklesseval.c used an incomplete copy of PyGenObject. Now they use PyGenObject directly.

https://bitbucket.org/stackless-dev/stackless/issues/111
(grafted from d9b92ad853297e83edc5a8310fd9016733a1ce6e)

Per pythia python#111 (4) + supervisor 05:59:04Z + theologian 05:59:52Z.
Reform addresses ritualization risk identified in pythia python#111: 2 retro-
frames already happened (meta-rule self-violation D-1777005990 + push-36
race-window citation rule); a third would calcify 'gate becomes ritual
that activates only when pythia surfaces the omission.'

Reform: cost-benefit + 8-incident-cite required AT AUTHORING of new
procedural rules, NOT retroactively. Retro-frames acknowledge violation
but do not satisfy the meta-rule. Authors who skip framing must self-
flag and provide framing in follow-up BEFORE rule treated as adopted.

Cost-benefit (per the meta-rule it codifies):
- Cost: ~0 procedural — discipline change only
- Benefit: blocks ritualization failure mode pythia python#111 (4) named
- 8-incident-cite: pythia python#105 'fever has name infection still spreads'
  framing — adding rules without addressing root cause repeats pattern;
  reform addresses root cause (ritualization) not surface (rule count)

Doc-only +2L (paragraph wrap; logical 6L per theologian count).

Authorization: theologian 05:59:52Z + supervisor 05:59:04Z (4) ADOPTED.
Atomic doc commit, expedited APPROVE per supervisor directive.