chore: revert bignumber.js patch for charts and reapply the original bignumber.js change to SQL editor in an opt-in fashion

[xtinec]

SUMMARY

This PR does the following:

• Revert "Fix rendering regression from the introduction of bignumber (Fix rendering regression from the introduction of bignumber #6937)"
• Consume the bignumber.js reversion by updating the version of superset-ui-connection
• Re-apply the bignumber conversion change (originally from Fix for #6590: Numeric values in columns sometimes returned as quoted strings  #6591) in actions/sqlLab.js (/results and /sql_json endpoints) where it is needed and make sure the existing unit tests for this still pass.
• Revert "Fix deck.gl form data (Fix deck.gl form data #6953)" b/c form data now returns snake_cased properties for deck vizzes.

BEFORE/AFTER SCREENSHOTS OR ANIMATED GIF

TEST PLAN

• Manually verified that bignumber.js related React runtime warnings in SQL editor (i.e. expecting a number but getting an object) disappeared
• Sanity check in charts and dashboards

ADDITIONAL INFORMATION

[ ] Has associated issue:
[ ] Changes UI
[ ] Requires DB Migration. Confirm DB Migration upgrade and downgrade tested.
[ ] Introduces new feature or API
[ ] Removes existing feature or API
[ ] Fixes bug
[x] Refactors code
[ ] Adds test(s)

REVIEWERS

@kristw @williaster @soboko
cc @betodealmeida @mistercrunch

[codecov-io]

Codecov Report

Merging #7210 into lyftga will decrease coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           lyftga   #7210      +/-   ##
=========================================
- Coverage    64.5%   64.5%   -0.01%     
=========================================
  Files         424     423       -1     
  Lines       20667   20651      -16     
  Branches     2268    2264       -4     
=========================================
- Hits        13332   13321      -11     
+ Misses       7212    7207       -5     
  Partials      123     123

Impacted Files	Coverage Δ
superset/assets/src/chart/ChartRenderer.jsx	7.57% <ø> (+0.53%)	⬆️
superset/assets/src/SqlLab/actions/sqlLab.js	70.64% <100%> (+0.29%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 200efac...76dc6f6. Read the comment docs.

[codecov-io]

Codecov Report

Merging #7210 into lyftga will decrease coverage by <.01%.
The diff coverage is 10.52%.

@@            Coverage Diff            @@
##           lyftga   #7210      +/-   ##
=========================================
- Coverage    64.5%   64.5%   -0.01%     
=========================================
  Files         424     423       -1     
  Lines       20667   20651      -16     
  Branches     2268    2264       -4     
=========================================
- Hits        13332   13321      -11     
+ Misses       7212    7207       -5     
  Partials      123     123

Impacted Files	Coverage Δ
...erset/assets/src/visualizations/deckgl/factory.jsx	0% <ø> (ø)	⬆️
superset/assets/src/visualizations/deckgl/utils.js	94.33% <ø> (ø)	⬆️
superset/assets/src/chart/ChartRenderer.jsx	7.57% <ø> (+0.53%)	⬆️
...c/visualizations/deckgl/layers/Scatter/Scatter.jsx	0% <0%> (ø)	⬆️
...t/assets/src/visualizations/deckgl/Multi/Multi.jsx	0% <0%> (ø)	⬆️
...c/visualizations/deckgl/layers/Geojson/Geojson.jsx	0% <0%> (ø)	⬆️
...ets/src/visualizations/deckgl/layers/Grid/Grid.jsx	0% <0%> (ø)	⬆️
...ets/src/visualizations/deckgl/layers/Path/Path.jsx	0% <0%> (ø)	⬆️
...sualizations/deckgl/CategoricalDeckGLContainer.jsx	0% <0%> (ø)	⬆️
...c/visualizations/deckgl/layers/Polygon/Polygon.jsx	0% <0%> (ø)	⬆️
... and 5 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 200efac...6b185fb. Read the comment docs.

[williaster]

I don't have as much context on the Deck.gl changes so probably good for someone else to review that. The big number + @superset-ui changes lgtm tho 👍

[mistercrunch]

I also have little context on this and this looks a bit scary. Quick reverts are cool, but long standing reverts are dangerous, I dug in and 95 PRs have been merged since #6937, some of these could have built upon assumptions on top of #6937 ... In general it may be easier (if possible) to push forward instead.

But if you feel like this is necessary and that you've covered all the basis, then I'm ok moving forward.

[xtinec]

@mistercrunch planning on pushing ahead. We're going to do a bug bash in staging with this change. There has been a lot of rendering regressions in charts and dashboards since the introduction of bignumber.js usage in superset-ui-connection in Jan. These regressions are in the core areas of the entirety of Superset. We discussed with the airbnb folks and decided we should do this reversion.

My thought is addressing the root cause of the past regressions outweighs the possibility of regressing the recent 95 PRs. We'll hopefully uncover any regression to these recent PRs in our bug bash.

[betodealmeida]

+1

Go for it, though I agree with Max, we should test some of the things that were added in between. I've had 2 fixes for the time filter that were related to this, eg.