Consider using smart pointers in ReSolve

[pelesh]

Based on @cameronrutherford comment: This is endemic to ExaGO and HiOp, but I would be curious if we could get away from managing raw pointers like this, and have more of our examples use more modern C++ resource management techniques with shared/unique pointers where absolutely necessary, and then just having non-pointer variables here that are cleaned up without delete.

Originally posted by @cameronrutherford in #83 (comment)

[cameronrutherford]

I take this as you agree with my suggestion, but this change will require a little bit of work :)

[pelesh]

I take this as you agree with my suggestion, but this change will require a little bit of work :)

We can do it in incremental steps. For vector/matrix data I think we need to keep raw pointers for portability. But for objects such as Gram-Schmidt, it would make more sense to use smart pointers.

CC @kswirydo