GeezerVison

`@vectorcall` accepted `*args`, but the generated code expected an
args tuple. Use fastcall form to parse the args array.

Co-Authored-By: Claude Opus 5.5 <[email protected]>

zip had no vectorcall, so every call built an args tuple and a kwargs
dict, and strict=True also went through PyArg_ParseTupleAndKeywords with
a temporary empty tuple.  Argument error messages are unchanged except
for unknown keywords: `zip(strict=1, x=2)` now reports "got an
unexpected keyword argument 'x'" instead of "takes at most 1 keyword
argument (2 given)".  The docstring gains a summary line;
`__text_signature__` is unchanged.

pyperf microbenchmark (PGO+LTO, run serially):

| Benchmark              | base    | zip                   |
|------------------------|:-------:|:---------------------:|
| zip(a, b, strict=True) | 118 ns  | 61.5 ns: 1.92x faster |
| zip(a, b)              | 64.9 ns | 53.1 ns: 1.22x faster |
| zip(a)                 | 55.0 ns | 47.1 ns: 1.17x faster |
| list(zip(a, b))        | 123 ns  | 113 ns: 1.09x faster  |
| zip()                  | 27.4 ns | 25.7 ns: 1.07x faster |
| Sub(a, b)              | 71.5 ns | 70.6 ns: 1.01x faster |

Benchmark hidden because not significant (1): zip.`__new__`

Benchmark code:

```python
import pyperf

runner = pyperf.Runner()
setup = "a = (1, 2, 3); b = [4, 5, 6]"
for stmt in ["zip()", "zip(a)", "zip(a, b)", "zip(a, b, strict=True)",
             "list(zip(a, b))"]:
    runner.timeit(stmt, stmt, setup)
runner.timeit('zip.`__new__`', 'new(zip, a, b)', setup + '; new = zip.__new__')
runner.timeit('Sub(a, b)', 'S(a, b)', setup + '\nclass S(zip): pass')
```

Co-Authored-By: Claude Opus 5.5 <[email protected]>

Co-authored-by: Stan Ulbrych <[email protected]>

StanFromIreland left a comment