Fix redis connection leakage

[hiimdoublej-swag]

Note: Before submitting this pull request, please review our contributing
guidelines.

Description

This PR is picked from #6985 since it has gone stale. (One of the maintainers asked for a rebase and that's all she wrote.)
This tries to resolve #6819 by ensuring there's only 1 redis connection pool across all the gevent concurrencies.

[codecov]

Codecov Report

Merging #7631 (8718c50) into master (aa9fd8a) will decrease coverage by 0.01%.
The diff coverage is 25.00%.

@@            Coverage Diff             @@
##           master    #7631      +/-   ##
==========================================
- Coverage   89.75%   89.74%   -0.02%     
==========================================
  Files         138      138              
  Lines       16996    17000       +4     
  Branches     2508     2509       +1     
==========================================
+ Hits        15255    15256       +1     
- Misses       1491     1493       +2     
- Partials      250      251       +1     

Flag	Coverage Δ
unittests	89.72% <25.00%> (-0.02%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
celery/app/base.py	95.92% <25.00%> (-0.53%)	⬇️

---

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 aa9fd8a...8718c50. Read the comment docs.

[auvipy]

would you mind adding some test for this change?

[auvipy]

redis integration tests are failing

[hiimdoublej-swag]

redis integration tests are failing

I take a look, thanks

[hiimdoublej-swag]

@auvipy How did you manage to run tests locally ? The build in /docker doesn't seem to work.
Complains about dependency conflicts.

[hiimdoublej-swag]

Well, I managed to get the tests running, but I think this fix introduced another problem with threading.

[auvipy]

yeah if you can fix that please push and let me know

[hiimdoublej-swag]

It seemed to stuck on result retrieval under threaded environments, not sure why.

[hiimdoublej-swag]

I'm starting to think that maybe this approach wasn't the correct fix, I'll close for now until I discover otherwise or found another fix.