Removes RS-LS truncation of stored columns

[KyleDavisSA]

Main changes of this PR

This PR removes the arbitrary truncation of all but the first 5 columns per time window for the least squares restart for IQ-IMVJ method. This negatively impacts the performance of the RS-LS method. The accumulation of columns is handled by the filtering the of the RS-LS columns during the restart step.

Motivation and additional information

Tests have shown that this truncation ignores information helpful for convergence. The problem of too many columns used is covered by the filtering step in

precice/src/acceleration/MVQNAcceleration.cpp

Lines 584 to 603 in cf1c06e

	impl::QRFactorization qr(_filter);
	qr.setGlobalRows(getLSSystemRows());
	// for QR2-filter, the QR-dec is computed in qr-applyFilter()
	if (_filter != Acceleration::QR2FILTER) {
	for (int i = 0; i < static_cast<int>(_matrixV_RSLS.cols()); i++) {
	Eigen::VectorXd v = _matrixV_RSLS.col(i);
	qr.pushBack(v); // same order as matrix V_RSLS
	}
	}
	// apply filter
	if (_filter != Acceleration::NOFILTER) {
	std::vector<int> delIndices(0);
	qr.applyFilter(_singularityLimit, delIndices, _matrixV_RSLS);
	// start with largest index (as V,W matrices are shrinked and shifted
	for (int i = delIndices.size() - 1; i >= 0; i--) {
	removeMatrixColumnRSLS(delIndices[i]);
	}
	PRECICE_ASSERT(_matrixV_RSLS.cols() == qr.cols(), _matrixV_RSLS.cols(), qr.cols());
	}

Author's checklist

• I added a changelog file with make changelog if there are user-observable changes since the last release.
• I ran make format to ensure everything is formatted correctly.
• I sticked to C++14 features.
• I sticked to CMake version 3.10.
• I squashed / am about to squash all commits that should be seen as one.

Reviewers' checklist

• Does the changelog entry make sense? Is it formatted correctly?
• Do you understand the code changes?

[uekerman]

Do we still need the member variable _usedColumnsPerTimeWindow then?
It's pretty suspicious that we can change sth like without any test failing. IMVJ is really not well tested.

[KyleDavisSA]

No, we would not need the variable _usedColumnsPerTimeWindow. It is weird that it is there at all, and I found the results between the different restart were better by removing this.

[uekerman]

We also need a changelog entry here.

[uekerman]

I assume that the original motivation to have this limit was efficiency of the restart, meaning that the restart itself does not become to expensive.