[MultiThreading] Introduce ParallelTetrahedronFEMForceField

[alxbilger]

Based on #3548

Similar to ParallelHexahedronFEMForceField.

addDForce and addKToMatrix are the parallelized functions.
addForce is not parallelized. It would require to duplicate code from the base class. Hence a refactoring is needed for a proper implementation (in another PR). Note that addForce is usually not the bottleneck.

Quick benchmarks show performances improvement on the provided example.

I think there is still room for improvement in addDForce. But it would require a few changes in the base class. Instead of having a force vector for each task, I suggest to store only the force on the DoFs that have been visited during the task. I am not sure if it will bring more perfs improvements.

In this PR, addKToMatrix is parallelized. It is not in ParallelHexahedronFEMForceField, but the same principle can be applied (in another PR).

---

By submitting this pull request, I acknowledge that
I have read, understand, and agree SOFA Developer Certificate of Origin (DCO).

---

Reviewers will merge this pull-request only if

• it builds with SUCCESS for all platforms on the CI.
• it does not generate new warnings.
• it does not generate new unit test failures.
• it does not generate new scene test failures.
• it does not break API compatibility.
• it is more than 1 week old (or has fast-merge label).

[alxbilger]

[ci-build][with-all-tests]

[alxbilger]

Benchmark on the provided example

Matrix-free CG:

Sequential:
[INFO] [BatchGUI] 1000 iterations done in 4.1899 s ( 238.669 FPS).
CG-Solve 3.21 ms (mean)

Parallel:
[INFO] [BatchGUI] 1000 iterations done in 2.82476 s ( 354.012 FPS).
CG-Solve 1.85 ms (mean)

Assembled CG

(CG with template="CompressedRowSparseMatrixMat3x3d"

Sequential:
[INFO] [BatchGUI] 1000 iterations done in 6.65858 s ( 150.182 FPS).
CG-setSystemMBKMatrix 2.34 ms (mean)

Parallel:
[INFO] [BatchGUI] 1000 iterations done in 4.1272 s ( 242.295 FPS).
CG-setSystemMBKMatrix 4.94 ms (mean)