Fix simultaneous Gröbner basis computation

[wegank]

This PR allows implementing dispatches for {initial_input,initial_gens,monomial,spair,hcm}_cmp using an immutable function. Note that this is the minimal breaking change in the context of #288.

[jerebertho]

Many thanks!
CI / FLINT had a lot of errors, which is very weird.

[wegank]

Last week, I've tested this PR on an ARM Neoverse N1 with 128 GB of RAM. Based on a set of 102 examples from https://gitlab.lip6.fr/eder/msolve-examples, the results show that the binary built from this PR achieved better performance in 53.9% of cases, with performance differences appearing only at the percentile level. Each example is solved using 8 threads and solely at the level of Gröbner bases; no real root isolation is performed.

[ederc]

Same behaviour for my tests.
@mohabsafey do you think that's ready to be merged?

[mohabsafey]

Yes, I think it is ready to be merged. Many thanks!

[ederc]

@wegank So I can move this PR from "draft" to "ready for review" and merge it, also from your side, right?

[ederc]

We will do a bugfix release first, then handle this PR.