Add radial background periodization

[krystophny]

Second split item of #175: the radial periodization of the background.
periodize_m (KIM/src/background_equilibrium/periodize_m.f90) provides the
C-infinity localizer weight and the periodized profile value for a tabulated
profile: untouched inside the resonant layer |r - r_m| <= dr_layer, blended
with its period-shifted copies in the transition zones, periodic with
L = 2 (dr_layer + dr_transition).

Stack: parent branch feat/flux-pumping/wp5-kernel-restore at 3959cf6c
(#177); this PR's diff is cumulative until the parent merges. Pipeline
integration behind a config switch is the fourth split item; nothing calls
the new module yet.

Invariant preserved

• The construction is the circulated prototype term-for-term: localizer
  weight exp(-2 pi/(1-t) exp(-sqrt(2)/t)) with clamped ends, wrap into
  [r_m - L/2, r_m + L/2), then the two shifted-copy blends.
• Verified symbolically before implementation (programme CAS record):
  the layer is left unmodified, the result is exactly L-periodic, values
  and derivatives through order 4 are continuous at the seam, and the
  localizer's derivatives vanish at both transition ends.
• Gradients must be taken of the periodized profile; periodizing a
  gradient differs in the transition zone (also part of the CAS record).
  The module documents that ordering constraint.
• Profile lookups reuse the repository's 4-point binsrc/plag_coeff
  Lagrange interpolation; the tabulation must cover three half-periods on
  each side of r_m, guarded by explicit errors.

Verification

Test fails on a degraded localizer (linear blend mutation)

$ sed -i 's|localizer_weight = exp(...)|localizer_weight = 1.0d0 - t|' periodize_m.f90
$ ctest --test-dir build -R test_periodize --output-on-failure
1/1 Test #25: test_periodize ...................***Failed    0.07 sec
 FAIL transition edge slopes agree
   got  =    82.996000000093773
   want =    3.4919999999374340

Test passes after restore

$ ctest --test-dir build -R test_periodize
1/1 Test #25: test_periodize ...................   Passed    0.06 sec

Full suite

$ ctest --test-dir build
100% tests passed, 0 tests failed out of 28

Checks: resonant layer unmodified (4-point Lagrange exact on the quadratic
profile, 1e-12 relative), exact L-periodicity across five periods, an
already-L-periodic profile passes through unchanged (1e-8), and one-sided
finite-difference slopes agree at the period seam and at the layer edge
(1e-3 relative, against an O(1) kink for a non-smooth blend).

Relation to Markus's Fokker-Planck periodic branch

This PR is an independent implementation of the C-infinity periodizer. Compare its seam, derivative, and untouched-layer behavior with feature/kim-forced-periodicity; do not retain two hidden periodizers. Multispecies preservation is specified in #186 and the final selection belongs to #191.