✨ chart-diff: add metadata-diff view, bug fixes

[lucasrodes]

• [x] New dataset: Plastic surgery dataset.
• [x] Bug fixed: Charts changed due to config not showing.
• [x] Feature: Compare metadata changes with some diff tool. ✨ Include GPT-summary, which only works locally (could we add OpenAI creds in staging servers?).
• [x] ~Feature: Compare data changes with some explorer-like tool.~ For a future PR.
• [x] Merge filtering options 'Chart modified / new' and 'Chart changes type' into one.
• [x] Change reject/approval options for metadata/data-triggered changes.
• [x] Change owidbot message.

BUG: Charts without no change type tag are not being shown.

• [x] DONE

Bug explained!

These charts are: new charts and charts whose config has changed for special fields. I fixed the bug for 'new charts'. For charts with only changes in fields specified by [configs_are_equal](https://github.com/owid/etl/blob/data/plastic-surgery/apps/wizard/app_pages/chart_diff/chart_diff.py#L240-L246) are not considered as different by this function. Hence, these are not tagged with the label `change_type='config'`. Consequently these charts are not shown unless we enable 'show all'. This, however, is not consistent with what [modified_charts_by_admin](https://github.com/owid/etl/blob/data/plastic-surgery/apps/chart_sync/cli.py#L492) from chart_sync is doing. It gets charts that have suffered any change in the config. And this is important, because the charts obtained by `modified_charts_by_admin` are those listed by owidbot. Consequently, in this PR we observe that there are **3 charts** listed by @owidbot, yet only two charts are shown by default in chart-diff. We should align these behaviours. Some ideas: 1. Further filter the charts obtained by `modified_charts_by_admin` by checking their configs and ignoring them based on the same criteria as `configs_are_equal`. 2. Change the logic from `configs_are_equal` and just tag a chart with _any_ change with `change_type='CONFIG'`. 3. Leave as-is, but make @owidbot ignore charts with changes in fields mentioned in `configs_are_equal`. We would have charts with such minor config changes in chart-diff, but wouldn't be shown. My personal preference is either 1 or 2. I find 3 confusing.

BUG: There are leaks and charts from other workflows are shown

likely fixed by https://github.com/owid/etl/pull/2797

[owidbot]

Quick links (staging server): Site	Admin	Wizard

Login: ssh owid@staging-site-data-plastic-surgery

chart-diff: ✅

• 2 new charts (2 reviewed)
• 1 modified charts (1 reviewed)

data-diff: ✅ No differences found

```diff + Dataset garden/health/2024-06-11/isaps_plastic_surgery + + Table isaps_plastic_surgery + + Column num_procedures Legend: +New ~Modified -Removed =Identical Details Hint: Run this locally with etl diff REMOTE data/ --include yourdataset --verbose --snippet ``` Automatically updated datasets matching _weekly_wildfires|excess_mortality|covid|fluid|flunet|country_profile|garden/ihme_gbd/2019/gbd_risk_ are not included

Edited: 2024-06-12 20:20:23 UTC Execution time: 3.84 seconds

[lucasrodes]

@Marigold FYI, I found couple of minor bugs. Tried to explain them in the PR description.

I estimate that you will start working like 2 hours before me, so I thought I'd ping you in case you want to look into this before I start working. Otherwise I'll work on it tomorrow.

[Marigold]

Thank you @lucasrodes. Just a heads up that I hotfixed two of those bugs on master & this branch, but there are other things to fix (or at least to discuss them). I'll write it up as soon as I finish firefighting.

[lucasrodes]

@Marigold I'd say you can play around with charts that originate from 'metadata' changes. If you run this locally, you will get gpt summaries (bc there are no credentials on staging to access openai gpt).

on this matter, do you think we could have credentials added to staging servers? or is it risky?

[lucasrodes]

Also, not really the need to review the new dataset pipeline.