Configure midpoint Krylov limits

[rogeriojorge]

Summary

• expose linear_restart and linear_max_restarts on ImplicitMidpoint
• pass those values directly to SOLVAX FGMRES
• exercise a reduced-basis solve in the midpoint integration test

Defaults remain 20, preserving the current solver behavior. This is stacked on #21.

Why

FGMRES allocates a shape-static Krylov basis, so restart controls compilation size and accelerator memory even when a solve converges early. Large plasma states that converge in one or two inner iterations should be able to select a smaller basis without reducing the nonlinear Newton iteration limit.

Validation

• clean-environment pytest -q (5 passed)
• 65x65 x 6x6x6, five midpoint steps: 1.079 s at 20/20 versus 0.832 s at 2/2
• compile/warm-up: 1.964 s versus 1.896 s
• saved Ck, Fk, and total energy were bitwise identical

The measured 23% runtime improvement is configuration-specific; the PR changes no default.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
spectrax/midpoint_solver.py	98.18% <100.00%> (+0.06%)	⬆️

🚀 New features to boost your workflow:

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