Effect of two-particles integrals screening tolerance factor

[jacharrym]

The default two-particle integrals screening value of 1e-10 is causing errors in CI calculations with large basis sets involving high-energy virtual orbitals.
Apparently, some integrals are wrongly discarded. This effect accumulates when doing the atomic-to-molecular orbital transformation, resulting in lower repulsion between two virtual orbitals, then CI loses variationality and boom! Explodes.

Notes to myself.

1. Add an input variable to select the integral screening tolerance
2. Test different values for this system

GEOMETRY
        e-(Be)  AUG-CC-PVQZ     0.00    0.00    -6.0000 addParticles=0
        e-(Be)  AUG-CC-PVQZ     0.00    0.00    6.0000 
        Be       dirac           0.00    0.00    -6.0000 
        Be     dirac       	0.00	0.00	6.0000
END GEOMETRY

TASKS
	method = "UHF"
        configurationInteractionLevel ="CISD"
END TASKS

CONTROL
        readCoefficients=F
        numberOfCIstates=1
        CIdiagonalizationMethod = "JADAMILU"
        overlapEigenThreshold=1E-10
END CONTROL

INPUT_CI
        species="E-ALPHA" core=2 active=0
        species="E-BETA" core=2 active=0
END INPUT_CI

3. Set a lower tolerance for CI