Add support for Python 3.14 and drop EOL 3.9

[hugovk]

Follow on from fsspec/s3fs#993.

[martindurant]

It is right to fix the warning in gcsfs (although the timeout would be useful!), or to get the CI job to pass, you can do the same as here for now.

[martindurant]

Can you comment on why strict=True needs to be added to zip(), but mode="r" (or "rt") needs to be removed from open() when these are both the default values? I would rather not have the big diff if possible.

[hugovk]

Can you comment on why strict=True needs to be added to zip()

It's a Ruff rule that triggers for 3.10: https://docs.astral.sh/ruff/rules/zip-without-explicit-strict/

  B905 `zip()` without an explicit `strict=` parameter
     --> fsspec/utils.py:570:21
      |
  568 |             for v in zip(
  569 |                 *sorted(
  570 |                     zip(paths, starts, ends),
      |                     ^^^^^^^^^^^^^^^^^^^^^^^^
  571 |                 )
  572 |             )
      |
  help: Add explicit value for parameter `strict=`

This is from pre-commit.

I can revert if you like? If so, we'll need to add B905 to the ignore list.

---

mode="r" (or "rt") needs to be removed from open() when these are both the default values?

This is UP015 from running ruff check --select UP --fix: https://docs.astral.sh/ruff/rules/redundant-open-modes/

I see UP015 is in the ignore list, I guess running with --select UP overrode it. There is a comment in the ignore list to "Fix these codes later", but happy to also revert it if you like?

[martindurant]

Yes, please make it so that strict= is not required, but the mode argument is allowed. The coder should have the freedom to decide what level of explicitness/verbosity is apropriate.

[hugovk]

Done!