Remove samples with tracking state = 2 for setting, not rising [0.5 d]

[Ian Culverwell]

https://trac.romsaf.org/ropp/changeset/5427/ropp_src/branches/dev/Share/dmi_trunk_9.0/ropp_io/ropp/ropp_io.f90 https://trac.romsaf.org/ropp/changeset/5427/ropp_src/branches/dev/Share/dmi_trunk_9.0/ropp_io/ropp/ropp_io_read.f90 https://trac.romsaf.org/ropp/changeset/5427/ropp_src/branches/dev/Share/dmi_trunk_9.0/ropp_io/ropp/ropp_io_read_ncdf_get.f90 https://trac.romsaf.org/ropp/changeset/5427/ropp_src/branches/dev/Share/dmi_trunk_9.0/ropp_io/tools/eum2bufr_ec.f90 https://trac.romsaf.org/ropp/changeset/5427/ropp_src/branches/dev/Share/dmi_trunk_9.0/ropp_io/tools/eum2ropp.f90 https://trac.romsaf.org/ropp/changeset/5427/ropp_src/branches/dev/Share/dmi_trunk_9.0/ropp_io/tools/ropp2bufr_ec.f90 (similar/consistent updates to ropp2bufr_mo.f90 has not been made since https://trac.romsaf.org/ropp/changeset/3430)

Original changesets:

Remove samples with tracking state = 2 for setting, not rising: https://trac.romsaf.org/ropp/changeset/5349/ropp_src/branches/dev/Share/dmi_trunk_8.1/ropp_io/ropp/ropp_io_read_ncdf_get.f90

[Ian Culverwell]

Potentially quite a big impact. Previously, when calculating LCF we had been testing whether the profile was rising with statements like

        IF (BTEST(data_cl%PCD, PCD_rising)) THEN    ! Rising occultation

But data_cl%PCD hadn't been calculated, but simply set to its default of 2^16^-1, which meant that all profiles were considered to be rising. Fortunately the ROPP test dataset data/eum_test.n4, is a rising one, so the change has no effect on that.

The other aspect of the change is a change to the poilcy for handling RS data with a tracking state of 2: from

! For rising profiles remove all data with tracking state = 2 in the 
! beginning of the record as these are not valid. 

to

! For setting occultations remove all data with tracking state = 2 at the
! beginning of the record as these are not all valid and may in the processing
! result in cutting off the whole rs/ol section. For rising occultations it is
! better to include them, since the processing will find out where to cut off the
! bad ones at the bottom and thereby keep the good ones with tracking state = 2.

Happy to be led by Stig on this.

Once again, there's no raw sampling data in our test dataset, so no impact on the automatic tests.

Commit change at r5944, and close ticket.