gh-157859: Fix cmath.atanh() returning inf at ±1 ± sqrt(DBL_MIN)*1j

[greg-const]

Fixes #157859.

Use the near-pole formula in cmath_atanh_impl() when
ay == CM_SQRT_DBL_MIN, as well as when it is smaller.

At the boundary, the general formula forms 4.0 / DBL_MIN, which
overflows even though the result is finite. Changing < to <= avoids
that intermediate overflow. This also fixes the corresponding
cmath.atan() inputs, since cmath.atan() uses the same implementation.

Add a regression test for both functions, covering all four sign
combinations and the immediately adjacent floating-point inputs. The test
fails without the C change and passes with it.

Testing

$ ./python -m test test_cmath -m testAtanhNearPoleBoundary
Result: SUCCESS   (FAILURE with the C change reverted)

$ ./python -m test test_cmath test_math test_complex test_float test_statistics test_builtin -j4
Total tests: run=772 skipped=17
Result: SUCCESS

$ ./python -m test -j4
Total tests: run=46,950 failures=1 skipped=2,584

The full test run reported failures in test_socket
(testInterruptedTimeout) and test.test_asyncio.test_ssl. Both passed
when rerun. testInterruptedTimeout also fails on an unpatched build in
the same environment, so these failures appear unrelated to this change.

$ make patchcheck

patchcheck passes apart from the expected NEWS entry added with blurb.

• Issue: cmath.atanh() returns infinity for complex(1.0, 2.0**-511) #157859

[python-cla-bot]

All commit authors signed the Contributor License Agreement.

[bedevere-app]

Most changes to Python require a NEWS entry. Add one using the blurb_it web app or the blurb command-line tool.

If this change has little impact on Python users, wait for a maintainer to apply the skip news label instead.