gh-140009: Improve performance of list_extend_dictitems by using PyTuple_FromArray

[eendebakpt]

Benchmark:

import pyperf runner = pyperf.Runner() setup=""" d ={ii: ii+1 for ii in range(200)} dct_items = d.items() """ runner.timeit(name="extend list with dict keys", stmt="l = []; l += d", setup=setup) runner.timeit(name="extend list with dict items", stmt="l = []; l += dct_items", setup=setup) 

Results:

extend list with dict items: Mean +- std dev: [main] 3.78 us +- 0.07 us -> [pr] 3.17 us +- 0.03 us: 1.19x faster Benchmark hidden because not significant (1): extend list with dict keys Geometric mean: 1.09x faster 

The benchmark is 19% faster (extend list with dict keys is a control benchmark).

• Issue: Improve performance by replacing PyTuple_Pack with PyTuple_FromArray #140009

[vstinner]

If PyTuple_Pack() is "inefficient", I would prefer to have a specialized PyTuple_Pack2() function (or PyTuple_MakePair()) rather than having to use the generic PyTuple_FromArray().