Stop k-eigenvalue inner sweeps at round-off

[bwhewe-13]

The k-eigenvalue inner loops exit when the sweep-to-sweep change drops below epsilon * 1e-3 (CG uses min(epsilon * 1e-2, 1e-9)). With a tight epsilon on a fine mesh that target is below what round-off allows, so the change levels off and every outer iteration runs all max_inner sweeps. On the 247k-cell IAEA quad mesh at epsilon=1e-12 this was over 10 minutes per outer; with the fix it is about 40 seconds.

detail::SweepFloor stops the sweeps once the change has stopped shrinking while already under 1e-6. The CG and 1-D loops stop after one non-decreasing sweep, the 2-D Gauss-Seidel loops after three since their change wobbles a bit more on the way down. The fixed-source and time-dependent convergence checks are left alone.

Tests: TestRoundOffFloor in test_cg_ab.py runs every k solver path (structured CG/GS, unstructured CG/GS on straight and skewed meshes, 1-D) at epsilon=1e-16 and checks there is no inner-solve warning. Before the fix the CG cases and skewed GS failed. Full suite passes, and k is unchanged to 12 digits at normal tolerances (IAEA CG at 1.25 cm, skewed IAEA triangles with GS, compared against 1.0.2).

[codecov-commenter]

Welcome to Codecov 🎉

Once you merge this PR into your default branch, you're all set! Codecov will compare coverage reports and display results in all future pull requests.

Thanks for integrating Codecov - We've got you covered ☂️