gh-140009: Replace PyTuple_Pack with PyTuple_FromArray in codeobject.c and _pickle.c

[RoryGlenn]

gh-140009: Replace PyTuple_Pack with PyTuple_FromArray

Summary

This PR replaces PyTuple_Pack calls with PyTuple_FromArray using stack-allocated arrays in two files:

• Objects/codeobject.c: 11 replacements in _PyCode_ConstantKey() — called for every constant in every code object during compilation (handles bool, bytes, float, complex, tuple, frozenset, slice, and fallback branches)
• Modules/_pickle.c: 3 replacements in fix_imports(), save_global(), and find_class() — dictionary key construction for protocol compatibility mappings

Motivation

PyTuple_FromArray avoids the variadic argument overhead of PyTuple_Pack by taking a C array directly. This follows the pattern established in PR #140010 (merged for Objects/listobject.c).

Pattern

// Before
key = PyTuple_Pack(2, type, op);

// After
PyObject *items[] = {type, op};
key = PyTuple_FromArray(items, 2);

Note: Py_TYPE(op) returns PyTypeObject *, so an explicit (PyObject *) cast is needed in the array initializers for codeobject.c.

Testing

• All compilation-related tests pass: test_code, test_compile, test_peepholer, test_symtable, test_codecs, test_marshal (743 tests)
• All pickle tests pass: test_pickle, test_pickletools (1,157 tests)
• Build completes with 0 warnings/errors

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

"

[python-cla-bot]

All commit authors signed the Contributor License Agreement.

[bkap123]

Thanks for the suggestions. As explained here #140009 (comment):

Note: the remaining cases of PyTuple_Pack(1, ...) and PyTuple_Pack(2, ...) are for the most part not performance critical. If opening a PR, please create proper benchmarks (using pyperf, hyperfine or some other benchmarking package) showing the performance gain for the change.

Since many of your changes deal with PyTuple_Pack(2, ...), I suggest that you include benchmarks to see if the changes are worthwhile.

[bkap123]

This is probably not necessary since nothing really changes from the user's perspective. Just mention that performance improved, similar to the NEWS entry for #140010