SBE timestamp guessing leads to invalid cast time

[ocehugo]

In profile mode, for a CTD cast read from a cnv file witch do not contain the time stamp of each sample, we have to guess the sampling time based on the startDate info, the scan index and the sampling frequency (fixed at 4hz).

The problem was the time stamp guessing function were not aware of offset scans within the cnv file, such as from up/downcast, multiple casts, or what the user has filtered out in the SBE processing software. The timestamp function always creates a full range vector with NaNs. Since the first time value is then picked as the cast date, this caused broken chain processing since metadata is filled with NaNs and wrong time variables.

Requirements:

• Improve the timestamp guessing to account for offset scans
• Update a wiki to recommend users to export the time in the SBE files.
• Regression tests