Ruff enable isort (with custom sections)

[tdadela]

Enable ruff isort-like import sorting & make code compatible with them.

Alternative approach to #2554.

[cyberw]

Awesome! Can you just have a look at my comments and maybe test run one or two of the example locustfiles to see that nothing broke there?

Then this is ready for merge...

[tdadela]

There are two independent problems to solve: circular imports and importing locust first in locustfiles.

1. Circular imports in locust/locust folder.
   (The most important file is locust/locust/__init__.py).
   Python executes imports (and other expression) from top to down a file.
   Let's consider the following scenario:
   After sorting imports in locust/locust folder (without adding if TYPE_CHECKING: and moving some imports here) the from .user.users import HttpUser, User statement of the locust/locust/__init__.py is located in line 25, below for example from .shape import LoadTestShape (line 21)
   Let's consider following the one-line script:

   from locust import LoadTestShape

   from locust import LoadTestShape makes interpreter try to import LoadTestShape from locust/locust/__init__.py → interpreter opens locust/locust/__init__.py and reaches line 21 → opens shape.py file → reaches from . import User line BEFORE definition of LoadTestShape → from . import User means "import User from file locust/locust/__init__.py"
   However, it's not possible to import User because User is in line 25 and interpreter currently process line 21.
   I see three solutions to this problem:

   1. Making from .user.users import HttpUser, User the first import after monkey.patch_all(). It's possible thanks to the use of # isort: skip:

   from .user.users import HttpUser, User  # isort: skip

   2. Adding if TYPE_CHECKING in some files.
   3. Using from .user.users import User instead of from . import User in some files.

2. In locustfiles Locust have to be imported first (due to gevent Monkey Patching).
   I see two solutions to this problem:

   1. Not sorting imports in locustfiles (like in Ruff enable isort #2554).
   2. Defining custom sorting section and custom selection order (see this PR – changes in pyproject.toml, documentation isort, documentation ruff).

[cyberw]

I see! I think option "1i" and "2ii" make the most sense - does that sound reasonable? Can you adjust your second PR? (or make a third one, if you prefer)

[cyberw]

Actually. Maybe the if TYPE_CHECKING makes more sense, as it shows the intent. I'll merge this as is, if you dont have anything more to add? (thanks for the explanations)

[tdadela]

i, ii, iii (and any combination of them) looks good to me.

[cyberw]

Huh. The build is failing after merge. Can you have a look? https://github.com/locustio/locust/actions/runs/7621650409/job/20758349498

[cyberw]

nvm, I fixed it myself

[tdadela]

(PR wasn't up-to-date with master branch. Some imports were added in the meantime.)