[HazenBabcock]

Ruff (--select F, the pyflakes rules) reports 486 findings. This fixes the
ones that mean "this raises if the path runs", plus the Python 2 code that
several of them turn out to be.

None of these paths is covered by the test suite, which is how they survived.

The crash class

where	what
simulator/pupil_math.py	sin_theta_2 = (n1/n2)*sin_theta_1 — neither name exists
multi_plane/fitting_mp.py	psf and j undefined, and "...c{1:d}".format(j) passes one argument for index 1
multi_plane/analysis_io.py	static_background used but never imported
admm/admm_lasso_c.py	def run(self, a_lamba, ...) against a body using a_lambda
multi_plane/mp_fit_arb_c.py	super(MPSFFnFit, ...) inside class MPPSFFnFit
simulator/emitters_on_lines.py	argls.nemitters for args.nemitters
jupyter_examples/multiplane_measure_psf.py	__main__ calls sCMOSSinglePlaneXML(); it is sCMOSSingleFrameXML()
test/test_drift_correction.py	__main__ calls _test_drift_correction_1(), renamed in #59

Plus two things defined twice: rolling_ball_lib_c.py imports numpy twice,
and spliner_configure_common.py imports spliner's settings module and then
shadows it with configure()'s parameter — which matters because
fista_decon/configure.py calls that function with its own settings.

Two of these needed a second fix the linter cannot see. pupil_math.py
also calls self.appylNARestriction() — an attribute, not a bare name, so
pyflakes' rules do not reach it, and fixing only the NameError moves the
failure four lines down. fitting_mp.py likewise would have gone from
NameError to IndexError on the format string. Both were caught only by
writing a test that actually calls the code.

Tests

test_fitting_mp.py is new — there was no test file for that module. It
builds the multiplane finder, turns check_mode on and reaches the block,
then checks the written TIFFs against a Gaussian PSF computed independently
in the test. The pupil_math test uses a geometry chosen so that 691 of 1024
pixels fall outside the NA; at the geometry the other tests in that file use,
nothing does, and the NA assertion would pass while testing nothing.

Python 2

pyproject.toml already declares requires-python = '>=3.10' and CI builds
3.10 through 3.12, so the 25 sys.version_info guards were dead code rather
than a supported path. Removed, along with the seven import sys statements
that became unused as a result, and the two Python 2 idioms in spline1D.py's
demo block.

Deliberately not done

• TSFProto_pb2.py keeps its two unicode calls. It is generated protobuf
  output marked "DO NOT EDIT", google.protobuf is not a dependency, and its
  two consumers use the Python 2 implicit relative import TSFProto_pb2.
  Those three files do not run at all; untangling them is a separate job.
  These are the only F821 findings left in the tree, down from 19.
• The other 462 findings, almost all unused imports and variables. Fixing
  them means rewriting files wholesale for no functional gain, which deserves
  its own decision rather than riding along here.

---

Full test suite passes, 280 tests. The four test files whose __main__
blocks changed were also run directly, since that is the path pytest never
executes and therefore never checked.