Currently, linsolve automagically infers the data type of its internal matrix operations and output data. We should provide a keyword for a user to override this decision if they, e.g., want to ensure float operations are used without having to check if every constant & data matrix they provide is of this type.