Fixed the ordering of the sm_rlzs

[micheles]

Only in this way full enumeration calculations can give the same mean curves as calculations with sampling. NB: this is changing the AELO numbers, but only for a single site in NAF:

# old
ASCE 7-16={'PGA': 0.52559, 'PGA_2_50': 0.52559, 'PGA_84th': 0.6275,
           'PGA_det': 0.6275, 'Ss': 0.75337, 'Ss_RT': 0.75337,
           'CRs': 0.9108, 'Ss_84th': 0.81386, 'Ss_det': 1.5,
           'Ss_seismicity': 'Moderately High', 'S1': 0.11795,
           'S1_RT': 0.11795, 'CR1': 0.92124, 'S1_84th': 0.17257,
           'S1_det': 0.6, 'S1_seismicity': 'Moderate'}
# new
ASCE 7-16={'PGA': 0.52559, 'PGA_2_50': 0.52559, 'PGA_84th': 0.62565,
           'PGA_det': 0.62565, 'Ss': 0.75337, 'Ss_RT': 0.75337,
           'CRs': 0.9108, 'Ss_84th': 0.8081, 'Ss_det': 1.5,
           'Ss_seismicity': 'Moderately High', 'S1': 0.11795,
           'S1_RT': 0.11795, 'CR1': 0.92124, 'S1_84th': 0.16942,
           'S1_det': 0.6, 'S1_seismicity': 'Moderate'}

The difference is really really small.