Optimize `PyFloat_Pack*()` functions using `_Py_bswap16/32/64()` functions

[vstinner]

I propose to optimize PyFloat_Pack*() C functions using _Py_bswap16/32/64() functions, and omit a copy to a temporary buffer in the native byte order.

Linked PRs

• gh-158567: Optimize PyFloat_Pack*() functions #158483
• gh-158567: Workaround GCC crash in PyFloat_Pack2() #158568

[vstinner]

Oh! gcc -O3 crashed on s390x Fedora Stable LTO 3.x when building floatobject.c!

during RTL pass: split1
Objects/floatobject.c: In function ‘PyFloat_Pack2’:
Objects/floatobject.c:2014:1: internal compiler error: Segmentation fault
 2014 | }
      | ^

Please submit a full bug report, with preprocessed source.
See <https://bugzilla.redhat.com/> for instructions.

I reported the GCC crash to GCC bug tracker as: https://bugzilla.redhat.com/show_bug.cgi?id=2544649.

[vstinner]

Update:

• GCC bug report: https://bugzilla.redhat.com/show_bug.cgi?id=2544649
• clang bug report: clang 22.1.8 miscompiles C code using _Float16 and __builtin_bswap16() on s390x llvm/llvm-project#228151
• Merged change gh-158567: Workaround GCC crash in PyFloat_Pack2() #158568 works around the GCC and clang bug

Since the compiler bug is worked around, I don't see anything left on the Python side. I close the issue.