fix AE potentials to remove unnecessary nonlocal channels

[camelto2]

the regularized ECPs for H,He,Li,Be where we smooth out the coulomb potential is purely a local potential. However, in order to get ppconvert to actually generate this potential from a GAMESS format (which requires at least one nonlocal channel and a local channel), you need to artificially add a "zero" nonlocal channel for S and make the P channel as local. The resulting xml from this approach also includes a nonlocal channel (which in turn creates a NLPP in qmcpack which we don't want for purely local potentials, since that means are wasting resources evaluating NLPPs for a "zero" potential in the "S" channel).

Since for a purely local potential we don't want this zeroed nonlocal channel, I remove it in the xml files as needed.

Addresses #70.

Note that the "cutoff" here is still 1e-10 which ppconvert is providing. However, since these xml files only consist of a local potential now, the cutoff is not used by qmcpack. For any local potential in qmcpack, the potential is splined on the grid provided in the xml file and doesn't use the cutoffs. For nonlocal potentials, the cutoff read in the xml file is used to define the grid over which the potentials are splined. Since there is now no longer any "nonlocal" channels, the cutoff parameter is irrelevant for these 4 xml files

[prckent]

How about putting 10 (or whatever the grid maximum is) for the cutoffs? This is less strange than 1e-10 and is at least related to another property in the file.

[camelto2]

How about putting 10 (or whatever the grid maximum is) for the cutoffs? This is less strange than 1e-10 and is at least related to another property in the file.

done

[prckent]

Thanks Cody