Refactor C++ and R interface for L2 regression functions (switch to double precision, native Rcpp exports)

[david20011999]

This pull request refactors several components of the ssemQr package to improve numerical precision, modernize the Rcpp interface, and enhance code maintainability and compilability.
The main updates replace single-precision (float) Eigen matrices with double-precision (double) equivalents, remove outdated .Call() wrappers, and rely on // [[Rcpp::export]] for cleaner R-C++ integration.

Details

1. R/solver.R

• Replaced the .Call("L2Regression", ...) interface with a direct call to the exported L2Regression() function.
• Simplified function call to ensure compatibility with the updated Rcpp interface.

2. R/utils.R

• Replaced .Call("ObjL2Regression", ...) and .Call("L2lamax", ...) by direct Rcpp exports.
• Ensured all L2 routines are now called via the modern Rcpp mechanism.

3. src/ssemQr.cpp

• Major refactor from MatrixXf (float) to MatrixXd (double) for better numerical precision.
• Replaced manual SEXP parsing with idiomatic Rcpp + Eigen type conversions.
• Added //[[Rcpp::export]] to L2Regression, ObjL2Regression, and L2lamax.
• Rewrote handling of list-type inputs using std::vector<ArrayXi> instead of std::vector<ArrayXd> for integer indexing.

src/LR.h

• Replaced all MatrixXf/VectorXf types with template-consistent MatType.
• Ensured consistent use of MatType in L2lr and L2lamax.
• Fixed identity matrix generation (MatType::Identity).

src/MatOp.h

• Unified index access using ix[i] instead of ix(i), improving compatibility with std::vector and ArrayXi.
• Changed iteration variable types (size_t) for safety.
• Updated centering function to use double and std::bind with std::minus<float> instead of deprecated bind2nd.

Testing

• Successfully compiled on both Linux (Ubuntu 22.04.5 LTS) and Windows 11 environments without errors or warnings.
• Confirmed that the Rmarkdown example included in the package documentation (model with simulated data) runs to completion without issues and reproduces expected results.