[WIP] Multi modular lifting container

[Breush]

• RNS Dixon solver (use Method::Dixon with multiModLifting = true) ;
• Hadamard bound changed to be exact value (so that ratrecon can work nicely on small matrices).

[jgdumas]

WIP as still failing with clang++;
and with g++ more with O0 than O2

[jgdumas]

test-hadamard-bound.C:46:38: error: ‘FastHadamardBound’ was not declared in this scope
46 | double fastHb = FastHadamardBound(A);
| ~~~~~~~~~~~~~~~~~^~~
t

[Breush]

• I added a static_assert to handle one bug when Modular<double> is not used.
• The other bug is reproducible with clang with test-last-invariant-factor -n 1 -b 3, but isn't here when OMP_NUM_THREADS=1 is set.

[Breush]

To make ./test-last-invariant-factor -n 5 -b 10; fail, do:

• have OMP_NUM_THREADS > 1
• disable optimisation flags on gcc or clang
• be on the rrzero branch of Givaro
• be on the Added firstprivate for iterator to TASK in FOR1D fflas-ffpack#280 branch of fflas-ffpack