update multigrid usage in test/mpi/solver

[yhmtsai]

This PR change the multigrid preconditioner setting in the test/mpi/solver
The original preconditioner criterion also check the absolute residual norm.
However, it will lead no update on the result from preconditioner when it meets the criterion already with initial input.
It makes the all thing stays the same in Cg from that point.
Thus, the previous CgWithMg will run until reaching iteration limit, which leads long time execution

It use the syntex from wiki cg

Assume the following $z_2$ is equal to $z_1$ when $z_1$ is initial guess and meet the criterion
$Mz_2 = r_2$
$\beta_1 = \frac{r_2'z_2}{r_1'z_1}$
$r_2'z_2 = (r_1-\alpha A p_1)'z_1 = r_1'z_1 - \frac{r_1'z_1}{p_1'Ap_1}(Ap_1)'z_1$
$= r_1'z_1 - \frac{r_1'z_1}{p_1'A(z_1+\beta_0p_0)}(p_1'A'z_1)$

$p_1'Ap_0 = 0$ and A is symmetric
Thus, we get it to

$r_1'z_1 -\frac{r_1'z_1}{p_1'Az_1}(p_1'Az_1) = r_1'z_1 - r_1'z_1 = 0$

Other than this, I also delete the unused part which is mentioned in #1736

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud