fix: update scan propagation for JAX 0.11's num_consts/num_carry removal

[gdalle]

Summary

• JAX 0.11.0 removed the num_consts/num_carry params on the scan eqn in favor of ft_in/ft_out (an FTTuple splitting invars into consts/carry/xs groups), which broke _prop_scan with KeyError: 'num_consts' (21 failing tests in tests/_interpret/test_scan.py).
• Derive num_consts/num_carry from eqn.params["ft_in"].unpack() instead.
• Bump the jax floor to >=0.11.0 to match. Since jax>=0.11.0 requires Python >=3.12, also drop Python 3.11 support (requires-python, classifiers, CI matrix).

Test plan

• uv run pytest — 4002 passed, 1 skipped, 0 failed
• uv run ruff check . / uv run ruff format --check .
• uv run ty check

🤖 Generated with Claude Code

[gdalle]

@adrhill why isn't CI running for this one?

[adrhill]

I think because you are not trying to merge into main, but into gd/badge (not sure what that branch is).

asdex/.github/workflows/ci.yml

Lines 1 to 7 in 3209417

	name: CI
	on:
	push:
	branches: [main]
	pull_request:
	branches: [main]

[adrhill]

Ah, it's the branch from PR #179.

[adrhill]

I merged it, I assume a follow-up commit will trigger CI now.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.81%. Comparing base (3209417) to head (d656b4f).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #180      +/-   ##
==========================================
- Coverage   93.81%   93.81%   -0.01%     
==========================================
  Files          61       61              
  Lines        4056     4055       -1     
==========================================
- Hits         3805     3804       -1     
  Misses        251      251              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[adrhill]

If this opens up new use-cases for scan, we should add tests (or at least open an issue documenting this for a follow-up PR, since it is undocumented).

[adrhill]

We will want to tag a patch release for this.

We could also upper bound JAX, but I worry uv might then fall back to the last non-upper-bounded version v0.5.1.

[adrhill]

Unlike Julia, the current jax>=0.9.0 doesn't avoid uv from installing v0.10.0 or even 1.0.0: https://packaging.python.org/en/latest/specifications/version-specifiers/#inclusive-ordered-comparison
For Julia-like behavior, we'd need ~=: https://packaging.python.org/en/latest/specifications/version-specifiers/#examples.

Touching on the NumPy version bound you mentioned in #181, I just found this:

Lower bounds are particularly critical when writing a library. It's important to declare the lowest version for each dependency that your library works with, and to validate that the bounds are correct — testing with --resolution lowest or --resolution lowest-direct. Otherwise, a user may receive an old, incompatible version of one of your library's dependencies and the library will fail with an unexpected error.

[adrhill]

Is this ready to merge @gdalle?

[gdalle]

Yes, up to you deciding on lower bounds