correction of index_function count_percentile_occurrences (tn10p and similar)

Test runs show that tn10p is OK with only small deviations wrt reference data. These devitions may be related to the problems listed below.

• Trivial: In setup.py move index function count_percentile_occurrences to its place in alphabetical order.
• tn90p, i.e. high percentile and > gives an early error cf. ~/PROJ/test_climix >more slurm-12739678.out
• txgt05p, i.e. low percentile and > produces the output file, but the data is not correct, cf. /nobackup/rossby20/sm_lbarr/PROJ/test_climix/txgt5p_EUR-11_ICHEC-EC-EARTH_rcp85_r12i1p1_CLMcom-CCLM4-8-17_v1_day_yr_19510101-21001231.nc. And there is a suspicious error message at the end of the log file, cf. ~/PROJ/test_climix >cat slurm-12742194.out
• In case of one percentile value as index function parameter; store it as an auxiliary coordinate. (How to handle the future possibility of several percentiles?).