Minor correction to dateclock.F and inert tracers metadata

[colucix]

In subroutine datevec the computation of seconds brings always to zero, because of

seconds=ABS(seconds-REAL(F_minutes*60,dp))

that should be

seconds=ABS(seconds-REAL(minutes*60,dp))

where minutes is the truncated (integer) value that shall be subtracted from seconds to obtain the correct number. This avoids for example duplicates in the output of diag.F when we the model is run with a DT that is less than one minute, i.e.

         1 2018-02-23 00:00:00.00  1.120707E-02  3.519684E+03  3.519696E+03  3.077123E+13
                     (255,070,30)  2.858101E-02  1.031344E-03  1.975004E-01  1.500445E+00
         2 2018-02-23 00:01:00.00  1.119368E-02  3.519694E+03  3.519706E+03  3.077129E+13
                     (136,003,30)  1.368026E-03  8.361375E-03  7.292163E+00  1.056991E+00
         3 2018-02-23 00:01:00.00  1.121212E-02  3.519704E+03  3.519715E+03  3.077135E+13
                     (136,006,30)  2.310436E-03  7.068164E-03  5.090048E+00  8.786054E-01
         4 2018-02-23 00:02:00.00  1.125206E-02  3.519712E+03  3.519723E+03  3.077142E+13
                     (135,008,30)  4.295297E-03  4.374091E-03  4.340897E+00  8.676624E-01
         5 2018-02-23 00:02:00.00  1.129371E-02  3.519719E+03  3.519730E+03  3.077148E+13
                     (108,007,30)  4.436244E-03  6.337892E-03  4.120553E+00  8.699228E-01

[arango]

Yes, good catch. I never run with very small timesteps, so I didn't noticed.

Also, I updated mod_ncparam.F and varinfo.yaml to include the metadata for the adjoint sensitivity of passive, inert tracers. Many thanks to Brian Powell for bringing this issue to my attention.