Do not set empty environment variables

[zc-devs]

Related #1911, #1913.

While #1913 fixes manual trigger and numbers/booleans for automatic triggers, it pollutes Pod's environment with empty strings:

        - name: CI_STEP_NAME
        - name: CI_WORKFLOW_NUMBER
          value: '1'
        - name: CI_COMMIT_TARGET_BRANCH
        - name: CI_COMMIT_REFSPEC
        - name: CI_STEP_NUMBER
          value: '0'
        - name: CI_COMMIT_PULL_REQUEST

[qwerty287]

Not sure, but I think there's a difference between "end var is empty" and "an env var is not set".

[zc-devs]

Before #1913, default values (empty string, zero, false) are filtered. It caused confusing, when I run pipeline manually. So, I removed it entirely. Now, there are a lot of empty string vars.
I think filtering zeros and falses is a mistake, because it is valid value. If nothing/not set needs to be expressed, null pointer should be used in the model.
However, it is a little bit different for strings, IMO. What's the meaning of empty string? The same as null, if we are talking about the model. Therefore this PR.

This PR should not break manual trigger, because manual envs is a different map, and this PR filters only empty strings in transition from model to envs.
Besides, looking at #1913 tests, there's no test with empty value. They are filtered in UI, as I remember correctly. So, the same logic, as in this PR.

[pat-s]

I can see arguments for both sides. I am not against a merge if this change gets properly documented, so that users can adjust their potential checks against empty/non-empty vars to now check for the existence.

While writing this, I realized that removing vars entirely makes checking a bit more cumbersome as one first needs to check for existence and then chain a check for a possible matching value.

@zc-devs We're currently curating issues for 3.0. If you want to get this into 3.0, please rebase and address the review in the next days.

[zc-devs]

I don't care, TBH.
Do maintainers want? Vote on this comment: 👍 want, 👎 don't.
It doesn't make sense to merge, correct and write the docs, if someone disagree.

[pat-s]

You opened this PR, so I guess you care at least a little.

Nobody objects and I am also fine with it, so it is (still) up to you to rebase and get it in :)

[woodpecker-bot]

Deployment of preview was torn down

[pat-s]

@zc-devs test failures