#4192 run_testsuite gives two failures when using SBCL 2.3.9

[Tomio Arisaka]

"run_testsuite (tests=["rtest_gamma"])" gives two failures when SBCL-2.3.8 or 2.3.9 is used to build Maxima.

$ rmaxima --norc
Maxima branch_5_47_base_424_g1d63a55ec_dirty https://maxima.sourceforge.io
using Lisp SBCL 2.3.9
Distributed under the GNU Public License. See the file COPYING.
Dedicated to the memory of William Schelter.
The function bug_report() provides bug reporting information.
(%i1) run_testsuite (tests=["rtest_gamma"]);
Testsuite run for SBCL 2.3.9:
Running tests in rtest_gamma: WARNING: redefining MAXIMA::SIMP-UNIT-STEP in DEFUN
WARNING: redefining MAXIMA::SIMP-POCHHAMMER in DEFUN

********************** Problem 384 (line 1846) ***************
Input:
relerror(gamma_incomplete(0.5, - 2.5 %i), 
                    - 0.5969141790423885 + 0.009214957317429537 %i, 3.0788e-15)


Result:
3.1371510054468076e-15

This differed from the expected result:
true

********************** Problem 390 (line 1882) ***************
Input:
relerror(gamma_incomplete(0.5, 2.5 %i), 
                    - 0.5969141790423885 - 0.009214957317429537 %i, 3.0788e-15)


Result:
3.1371510054468076e-15

This differed from the expected result:
true

825/827 tests passed

The following 2 problems failed: (384 390)

Updated "sb-kernel:two-arg-/" of SBCL causes this issue.
"sb-kernel:two-arg-/" is the SBCL function which computes the division between two complex numbers.

But I don't think this issue is a bug of SBCL.
Because "sb-kernel:two-arg-/" has been optimized since SBCL-2.3.8.
So the result of "sb-kernel:two-arg-/" is a very small difference between old versions of SBCL and new one.
Therefore, the result of gamma_incomplete is a very small difference, too.

See next commits:
https://sourceforge.net/p/sbcl/sbcl/ci/a07cacff6c9f0c9e79d5dc5bdd7eba70bbfaf781/
https://sourceforge.net/p/sbcl/sbcl/ci/f3f8e339c6690c718096ca0443a036a5f3a95e8a/