update Singular_jll and libsingular_julia_jll

[hannes14]

address comments to #722

[fingolfin]

Hmm, there is a CI failure in the OscarCI tests; the output of an absolute_primary_decomposition invocation changed. It was so far:

│ 2-element Vector{Tuple{MPolyIdeal{QQMPolyRingElem}, MPolyIdeal{QQMPolyRingElem}, MPolyIdeal{AbstractAlgebra.Generic.MPoly{nf_elem}}, Int64}}:
│  (ideal(z^2 + 1, y + 1), ideal(z^2 + 1, y + 1), ideal(z - _a, y + 1), 2)
│  (ideal(z^6 + 4*z^3 + 4, y - z^2), ideal(z^3 + 2, y - z^2), ideal(z - _a, y - _a^2), 3)

The final _a^2 now is _a*z....

Is this expected? In particular: Is this a regression? Or a bugfix? Or just randomness?

[hannes14]

Singular_jll has a changed absolute_primary_decomposition, in order to address Singular/Singular#1186.
But the ideal in question (z-_a,y-_a*z) is the same as (z-_a,y-a^2) as z=_a here, so both (the old and the new result) are correct.
But why reports OscarCI with julia 1.6 no diffrences?