Consistent way to deal with Raffenetti type integrals

[leethomo86]

At the moment Raffenetti 1, 2 and 3 integrals are read off disk separately, even though they are frequently used together. This can lead to inconsistencies where a vector of MQC_2ERIs type have to passed to a subroutine. It would be better to deal with Raffenetti types automatically within the MQC_2ERIs procedures, but then we might lose flexibility that we get treating Raffenetti types separately. Anyway, I propose changing the code so that there is different Raffenetti types are stored in MQC_2ERIs as with do for MO 2ERIs. Any thoughts?