Migrate linting from black to ruff, with a stricter rule set enforced by pre-commit

[AlexLipp]

The repo currently lints with black --line-length=100 only, pinned at black==22.6.0 in .pre-commit-config.yaml. That is formatting alone — no import sorting, no unused-import detection, no modernisation, no bug-pattern checks. Proposal: replace it with ruff, which covers formatting and linting in one fast tool.

Proposed configuration

Add to pyproject.toml (which currently has only [build-system] and [tool.pytest.ini_options]):

[tool.ruff]
line-length = 100          # keep the repo's existing width
target-version = "py311"

[tool.ruff.lint]
select = [
  "E", "F",      # pycodestyle + pyflakes
  "I",           # isort - import order
  "UP",          # pyupgrade - modern syntax
  "B",           # bugbear - common bug patterns
  "SIM",         # simplify - redundant code
  "C4",          # comprehension cleanups
  "NPY",         # numpy-specific gotchas
  "PTH",         # use pathlib over os.path
  "RUF",         # ruff's own extra checks
]

And replace the black hook in .pre-commit-config.yaml:

repos:
  - repo: https://github.com/astral-sh/ruff-pre-commit
    rev: v0.16.0
    hooks:
      - id: ruff
        args: [--fix]
      - id: ruff-format
  - repo: https://github.com/pre-commit/pre-commit-hooks
    rev: v5.0.0
    hooks:
      - id: trailing-whitespace
      - id: end-of-file-fixer
      - id: check-added-large-files

Refactoring the codebase to conform

This is the bulk of the work, and the reason it needs to be a deliberate pass rather than a drive-by:

• network_unmixer.py is entirely under # fmt: off (line 2), so no formatter has ever touched it. It has many lines well over 100 characters (one docstring line is 744 characters). Deciding whether to lift # fmt: off — and reflowing if so — is the single biggest decision here. ELEMENT_LIST carries # fmt: skip and must keep it.
• Expect a meaningful number of UP and PTH hits: the codebase uses typing.Dict/List/Optional throughout (pre-PEP 585/604 style) and os.path in places.
• B will likely flag the bare assert statements used for input validation, and the broad raise Exception(...) calls in solve (TRY/EM rules would too, if adopted).
• The codebase is annotated for pyre and carries deliberate # pyre-fixme comments — these must survive the pass.
• Any rule that genuinely needs suppressing should get a # noqa: RULE with a one-line reason, not a blanket ignore.

Suggested sequencing

1. Add the ruff config and run ruff check in report-only mode to size the problem.
2. Land the mechanical --fix-able changes first (imports, comprehensions, pyupgrade) in one commit.
3. Handle # fmt: off in network_unmixer.py as a separate, reviewable commit.
4. Address remaining hand-fix rules.
5. Swap the pre-commit hooks and re-run pre-commit install; note in the README that collaborators need to run it once after cloning.

Keeping hooks to lint/format only — the full test suite is slow by design (Hypothesis, deadline=None) and belongs in CI, not a per-commit hook.