Compress original least-squares rows by streaming Givens rotations.
The returned upper design and transformed right side preserve the variable
dependent objective; the remaining orthogonal residual is a constant term.
No normal matrix, pivot-based rank claim or inverse is formed. The caller
must qualify the compressed original design independently before solving.
Compress original least-squares rows by streaming Givens rotations. The returned upper design and transformed right side preserve the variable dependent objective; the remaining orthogonal residual is a constant term. No normal matrix, pivot-based rank claim or inverse is formed. The caller must qualify the compressed original design independently before solving.