feat: Lint with ruff

[dmartinol]

What this PR does / why we need it:
Replace the current linting and formatting stack with ruff>=0.3.3 (latest version).

Collateral changes:

• Update a few source code files to adapt to the known differences between isort and ruff, see here.
• Remove sdk/python/setup.cfg because not supported by ruff, and replaced by the equivalent pyproject.toml.
  • Move here also the configuration for mypy .
• Remove lint category B' as it seems to be never raised by flake8` (https://flake8.pycqa.org/en/latest/user/error-codes.html)
  • Instead, this category raises numerous errors with ruff (flake8-bugbear)
• Remove lint category C, because the only error raised by that rule was C901 (here, which was also added to the ignore list.
  • Vice versa, ruff has other errors in the category C4 (see here that would be raised instead.
  • Note about code complexity: flake8 and ruff compute different complexities for our code base, so it would be difficult to map the max-complexity number to a value that matches the new linting tool.
• Added ignore rule for E721, a new error code introduced by flake8>=6.1.0, the reference version for ruff
  • ruff raises a few violations in our codebase for this error, which could be fixed in a dedicated PR
  • Final note: the list of files and lines of code affected by this error is different in flake8 version 7.0.0 and ruff, meaning that ruff re-implemented this rule with a different behavior
• Remove lint category T4 because never raised by flake8 and unknown to ruff (Unknown rule selector: T4).
• Remove ignore rule for W503 because never raised by flake8 and unknown to ruff (Unknown rule selector: W503).
• Add rule I representing the isort category (see here)
• Update requirements files using pip-compile.

Which issue(s) this PR fixes:
Fixes #3956

[sudohainguyen]

ruff check is not enough isn't it?
there's still ruff format --check

[dmartinol]

ruff check is not enough isn't it?

You mean instead of python -m ruff check feast tests?
1- I just wanted to preserve the existing convention used in the Makefile to run modules with python -m
2- If agreed, we can change to ruff check feast tests (ruff check alone scans also undesired folders like docs)

there's still ruff format --check
Isn't it better to separate the concerns and use check for linting and format only for formatting, as in python -m ruff format feast/ tests/?

Note: I see that the latter command tries to format more files than the ruff check --fix, so I think I need to investigate a bit more the formatting options that were previously implemented by black --target-version py38 feast tests

[dmartinol]

Adding ruff formatting to replace black and the 52 files updated due to this change.
All changes seems due to the Known Deviations from Black.

[shuchu]

by any chance, will you update the Document regarding the code style & lint? or that will be a separate PR. https://docs.feast.dev/project/development-guide#code-style--linting

[dmartinol]

by any chance, will you update the Document regarding the code style & lint? or that will be a separate PR. https://docs.feast.dev/project/development-guide#code-style--linting

Yes it's here

I have some doubts about the sentence "Conforms to Black code style" because even if it's implemented by ruff format, the ruff formatter page states:

The Ruff formatter is an extremely fast Python code formatter designed as a drop-in replacement for Black, available as part of the ruff CLI via ruff format.

[jeremyary]

@sudohainguyen @shuchu @tokoko any further thoughts to share on the topic here? Just hoping to see if there's more for us to tackle next week. Thanks!

[sudohainguyen]

ltgm

[tokoko]

lgtm