FEAT: logaddexp2 ufunc implementation

[SwayamInSync]

Copilot Summary

This pull request adds full support for the logaddexp2 operation in quad-precision and long double types, including its implementation, exposure as a NumPy ufunc, and comprehensive tests for correctness and special cases. The release tracker is also updated to reflect this new support.

Implementation of logaddexp2 operation:

• Added a numerically stable implementation of quad_logaddexp2 for quad-precision (Sleef_quad) and ld_logaddexp2 for long double, handling all special cases (NaN, infinities, etc.). (quaddtype/numpy_quaddtype/src/ops.hpp) [1]], [2]])

Integration with NumPy:

• Registered the new logaddexp2 operation as a universal function (ufunc) for both quad and long double types, making it available in the NumPy interface. (quaddtype/numpy_quaddtype/src/umath/binary_ops.cpp) ([quaddtype/numpy_quaddtype/src/umath/binary_ops.cppR246-R248])

Testing improvements:

• Added comprehensive parameterized tests for logaddexp2, covering a wide range of values (including edge cases like NaN and infinities), mathematical properties, and its relationship to logaddexp. (quaddtype/tests/test_quaddtype.py) ([quaddtype/tests/test_quaddtype.pyR531-R637])

Documentation and tracking:

• Updated the release tracker to indicate that logaddexp2 is now fully supported for both quad and long double types. (quaddtype/release_tracker.md) ([quaddtype/release_tracker.mdL14-R14])

[juntyr]

LGMT

[SwayamInSync]

Since true_divide ufunc uses the ufunc divide <ufunc 'divide'> which is already registered so just adding its test cases here

[SwayamInSync]

Lets take this in, maybe job-14 got stuck