Kernel specialization of EwBinaryObjSca for DenseMatrix,double and double scalar multiplication

[aristotelis96]

Currently EwBinaryObjSca-kernel supports element-wise multiplications between dense matrices and scalars of the same value type (DenseMatrix<double> * double, DenseMatrix<int> * int). We need to support this operation between different value types as well. This will allow us to run more scripts like scripts/lmDS.daph, which requires an element-wise multiplication between DenseMatrix<double> and an int.

We need to specialize kernel EwBinaryObjSca for DenseMatrix<double> * int. The result should be a DenseMatrix<double> matrix.

This issue asks you to implement

1. A partial specialization of the EwBinaryObjSca-kernel for DenseMatrix<double> * int64_t (result DenseMatrix<double>).
2. Ideally some test cases in test/runtime/local/kernels/EwBinaryObjScaTest.cpp
3. Optionally, partial specilization for DenseMatrix<int64_t> * double (result DenseMatrix<double>).

[aristotelis96]

@pdamme do we still need this? I remember in #402 we discussed that type inference + type promotion solve these kind of issues, however, I think it depends on whether an operation has trait CastArgsToResType. Is there a benefit having different specializations at the kernel-level (and removing the trait)? In my opinion, if a kernel specialization is offering a faster/better implementation then the anwser is probably yes, though I am not sure if this applies to this case..