[LinearSolver.Direct] EigenDirectSparseSolver: Fix out-of-bounds read on matrices with empty rows

[fredroy]

advised by Claude while doing

• [LinearSystem] MatrixProjectionMethod: Fix out-of-bounds matrix writes #6256

its point:

copyNonZeros() does not store the rows which are entirely zero, so the filtered matrix held fewer row offsets than the Eigen::Map built from it reads, and the factorization accessed memory past the CRS arrays. fullRows() now restores those rows before the mapping.

Any remaining mismatch means the matrix contains entries outside its declared size ‚ CompressedRowSparseMatrix grows silently to store them , which is now reported instead of being mapped, the solver being left invalid so that solve() does not use an unfactorized Eigen solver.

I also instructed him to make unit tests to cover this case.

[with-all-tests]

---

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]

I located a function in CompressedRowSparseMatrixGeneric.h that could be widely used wherever a CRS is manipulated, including in the solver of this PR: check_matrix. Currently, it's used nowhere. The problem is that this function implementation is suspect, there is no test and no documentation. I suggest to have a common tool (such as this function) to avoid testing code wherever a matrix is used.