Follow-up from "Unify normalized data columns logic in Ci::JobDefinition"

The following discussion from !207802 (merged) should be addressed:

• @lma-git started a discussion: (+5 comments)

  @fabiopitino @tianwenchen @mbobin

  🤔 I'm not sure if this has already been discussed, but I just noticed that we're generating the checksum before we set the normalized column default values. Couldn't this be a problem if we decide to change a column default one day? I mean, it would be very unlikely that we'd change them, but it's still possible, especially if we might have more than just interruptible one day.

  In such scenario, we could have a new job pointing to a job_definition with a matching checksum but an old (incorrect) default value.

  I think it would be safer for us to generate the checksum after setting default values so we can capture the exact state of all attributes. (Edit: To clarify, I mean to also include it into the hash on which we generate the checksum.)

  Wdyt?