sagemath: update to 10.9.

[tornaria]

• ecl: update to 26.5.5.
• maxima: update to 5.49.0.
• lcalc: update to 2.2.1.
• libbraiding: update to 1.3.2.
• libhomfly: update to 1.04.
• libmpc: update to 1.4.1.
• python3-cypari2: add fix for memleak
• python3-mpmath: update to 1.4.1.
• python3-sympy: patch for mpmath 1.4
• sagemath: update to 10.9.

Testing the changes

• I tested the changes in this PR: YES|briefly|NO

[tornaria]

@dkwo this is a draft, let's test a bit before merging. Is there anything else that would be useful to update at the same time?

[dkwo]

On aarch64-gnu, I see two test failures during do_check:

python3 -m sage.doctest --warn-long 10.0 --random-seed=21274428239306805655443364373838102539 sage/matrix/matrix_mod2_dense.pyx
    Killed due to segmentation fault

which runs fine from installed sage, and

python3 -m sage.doctest --warn-long 10.0 --random-seed=21274428239306805655443364373838102539 sage/matrix/matrix_double_dense.pyx
**********************************************************************
File "sage/matrix/matrix_double_dense.pyx", line 1926, in sage.matrix.matrix_double_dense.Matrix_double_dense.SVD
Failed example:
    U*S*V.transpose()  # tol 1e-15
Expected:
    [0.9999999999999996 1.9999999999999998]
    [               3.0 3.9999999999999996]
    [ 4.999999999999999  6.000000000000001]
Got:
    [0.9999999999999996 1.9999999999999978]
    [2.9999999999999996  3.999999999999999]
    [               5.0                6.0]
Tolerance exceeded in 1 of 6:
    1.9999999999999998 vs 1.9999999999999978, tolerance 2e-15 > 1e-15
**********************************************************************
1 item had failures:
   1 of  24 in sage.matrix.matrix_double_dense.Matrix_double_dense.SVD
    [612 tests, 1 failure, 0.52s wall]

which gives the same (strange) result also with installed sage.
Overall, this looks quite good to me!

[tornaria]

The segfault is something that happens occasionally at random in every arch. The other one, we can change the tolerance from 1e-15 to 1e-14 and forget about it.

There are failures in i686 but I think upstream has given up on 32 bit 🤷