WIP - Don't do jobs if station RMS is "bad"

[ThomasLecocq]

here I summarise the discussions/code done with @awatlet today to avoid using toppled over seismic nodes in the CC calculations

1. scan archive
2. compute QC (psd)
   2.1. extract RMS
   2.2 define bad periods, save to csv
   2.3 use csv to delete CC jobs
3. back to normal process: compute_cc, etc...

[ThomasLecocq]

temporarily here, the job cleaner is (and its CSV):

linecol,start,end
2L_01103,18/11/2018 00:00,04/12/2018 00:00
2L_01104,08/08/2018 15:00,14/08/2018 00:00
2L_01104,18/11/2018 21:00,04/12/2018 00:00
2L_01105,16/11/2018 22:00,28/11/2018 15:00
2L_02103,25/11/2018 20:00,27/11/2018 21:00
2L_02104,13/11/2018 22:00,04/12/2018 00:00
2L_02105,25/11/2018 23:00,28/11/2018 10:00
2L_03102,08/08/2018 19:00,14/08/2018 00:00
2L_03103,25/11/2018 20:00,27/11/2018 13:00
2L_03104,02/08/2018 20:00,14/08/2018 00:00
2L_03104,19/11/2018 20:00,04/12/2018 00:00

"""
Job cleaner
"""

bad = pd.read_csv("data_check.csv", parse_dates=["start", "end"], dayfirst=True)
print(bad.head())
c = 0
for sta, group in bad.groupby("linecol"):
    print(sta, len(group), "bad periods")
    for id, period in group.iterrows():
        cmd = "DELETE FROM jobs where pair like '%%%s%%' and day BETWEEN '%s' and '%s' " % (sta.replace("_","."),period.start.date(), period.end.date())
        os.system("msnoise db execute \"%s\"" % cmd)