The new Matrix.inverse() can be used to factor the current code

[mbaudin47]

What is the idea?

The PR 2597 introduced new inversion features for a general or symmetric Matrix.

This feature can now be used to factor some existing code.

• QRMethod::getGramInverse() : QRMethod.cxx#L187
• QRMethod::getGramInverseDiag() : QRMethod.cxx#L199
• CholeskyMethod::getGramInverse() : CholeskyMethod.cxx#L290
• CholeskyMethod::getHDiag() : CholeskyMethod.cxx#L307
• DickeyFullerTest::estimateDriftAndLinearTrendModel() : DickeyFullerTest.cxx#L188
• EllipticalDistribution::update() : EllipticalDistribution.cxx#L486
• GeneralizedExtremeValueFactory::buildCovariates() : GeneralizedExtremeValueFactory.cxx#L977
• JointDistribution::getIsoProbabilisticTransformation() : JointDistribution.cxx#L1108
• RandomMixture::setDistributionCollectionAndWeights() : RandomMixture.cxx#L495
• GeneralizedParetoFactory::buildMethodOfLikelihoodMaximizationEstimatorWithStartingPoint() : GeneralizedParetoFactory.cxx#L450
• DistributionImplementation::getInverseCholesky() : DistributionImplementation.cxx#L3346

This review shows that a new inverse() method should be implemented for a triangular matrix.

• LinearModelStepwiseAlgorithm::computeLogLikelihood() : LinearModelStepwiseAlgorithm.cxx#L605

Furthermore, a new inverseInPlace() method should be implemented for some purposes:

• InverseWishart::setV() : InverseWishart.cxx#L352

Why is this needed?

This will simplify the source code and clarify the algorithm.

Additional Context

No response