Only send 1 event per prediction & fix tests

[hmstepanek]

Overview

1. This changes from sending 1 event per feature/label value to 1 event per prediction (Also referred to as a change in "schema").
   • Replace feature_name & feature_value and label_name & label_value to feature.<name> = <value> and label.<name> = <value>.
   • Remove feature_type and label_type.
   • Add a new schema attribute.
2. Remove pypy38 from test suite and remove compile dependencies from Dockerfile. Scipy wasn't able to compile consistently regardless of installing the necessary dependencies and adding these extra dependencies broke the kafka compile.

[github-actions]

🦙 MegaLinter status: ✅ SUCCESS

Descriptor	Linter	Files	Fixed	Errors	Elapsed time
✅ PYTHON	bandit	1		0	6.32s
✅ PYTHON	black	3	0	0	1.08s
✅ PYTHON	flake8	3		0	0.54s
✅ PYTHON	isort	3	0	0	0.3s
✅ PYTHON	pylint	3		0	4.31s
✅ YAML	prettier	1	1	0	0.98s
✅ YAML	v8r	1		0	2.13s
✅ YAML	yamllint	1		0	0.37s

See detailed report in MegaLinter reports
Set VALIDATE_ALL_CODEBASE: true in mega-linter.yml to validate all sources, not only the diff

MegaLinter is graciously provided by

[codecov-commenter]

Codecov Report

❗ No coverage uploaded for pull request base (develop-scikitlearn@c77adce). Click here to learn what that means.
The diff coverage is n/a.

@@                  Coverage Diff                   @@
##             develop-scikitlearn     #867   +/-   ##
======================================================
  Coverage                       ?   81.66%           
======================================================
  Files                          ?      189           
  Lines                          ?    19401           
  Branches                       ?     3341           
======================================================
  Hits                           ?    15843           
  Misses                         ?     2631           
  Partials                       ?      927           

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[TimPansino]

To be clear, we're removing PyPy3.8 testing because we can't get scipy to compile, yes? And these dependencies that are being dropped were added for that?

[TimPansino]

Ah I didn't read the second half of the PR description, I see you already covered that

[TimPansino]

Have we tested that this works for string return vals? Seems like it might split something like "abc" to ["a", "b", "c"].

[hmstepanek]

Very good point! For the label columns, the values are only numeric (these values come from the output of the model prediction) so we should not need to handle this scenario.

[TimPansino]

Overall looks good to me other than the 1 question. Have we gotten this validated in the UI as well?

[hmstepanek]

It has been validated in the UI. I'm checking on that string related question now...