fix: crash if requested output field for inference doesn't exist in dataset

[ividal]

What's changing

If a field is requested for outputs and it doesn't exist in the input dataset (which is to be expected), inference crashes (and, by extension, so does annotation).

This PR:

• copies every content from the original dataset to the new one that should be stored (so the user can always delete the previous one if they prefer)
• creates or overwrites the desired column. This changes the previous behaviour of keeping whatever was in the existing column (which meant we did inference for nothing if there were already predictions/annotations there).

Closes #585

How to test it

Steps to test the changes:

1. Upload the csv listed in the issue - keep the dataset id
2. Launch via API an inference job with any model, using the dataset id
3. The inference job should succeed.

Additional notes for reviewers

This changes the previous behaviour of keeping whatever was in the existing column (which meant we did inference for nothing if there were already predictions/annotations there).

I already...

• Tested the changes in a working environment to ensure they work as expected
• Added some tests for any new functionality
• Updated the documentation (both comments in code and product documentation under /docs)
• Checked if a (backend) DB migration step was required and included it if required

[ividal]

⚠️ Rebased.

[aittalam]

TY Irina for this PR!

I left a couple of minor comments but once they are addressed (and code from main is merged again 😅) this LGTM 🙏

[javiermtorres]

LGTM