Cap the k-space interpolation stencil at one period

[AdamField118]

Fixes #81.

Caps the k-space interpolation stencil at one period, using the closed-form
Dirichlet kernel for the sinc interpolant. The k-image is periodic and indices are
already wrapped with % nkx, so aliases past one period land on points the stencil
already covers, sinc's ixrange is 63662 at the default kvalue_accuracy, so a
9x9 draw asked for ~3e11 elements and failed to allocate. It now runs in 0.05s.

Compact interpolants (Nearest, Linear, Cubic, Quintic, Lanczos 3/5) are bitwise
identical before and after. With sinc, kValue now agrees with Quintic and
Lanczos(7) on the same image.

One deliberate difference from GalSim: the fast path triggers on 2*xrange <= n
rather than xrange() <= n, since the single-alias shortcut is only valid when the
nearest other alias falls outside xrange (the looser test is off by ~1.7e-2 for
Quintic at n=4). xvalWrapped has no call sites in GalSim, so it isn't reachable there.

Re-enabled SincInterpolant() in test_interpolatedimage_utils.py, where it was
commented out.

[codspeed-hq]

Merging this PR will improve performance by 57.46%

⚠️ Unknown Walltime execution environment detected

Using the Walltime instrument on standard Hosted Runners will lead to inconsistent data.

For the most accurate results, we recommend using CodSpeed Macro Runners: bare-metal machines fine-tuned for performance measurement consistency.

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

⚡ 12 improved benchmarks
✅ 26 untouched benchmarks

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	WallTime	test_benchmark_moffat_init[run]	86.8 µs	37.5 µs	×2.3
⚡	WallTime	test_benchmarks_lanczos_interp[xval-no_conserve_dc-run]	123.6 µs	63.5 µs	+94.5%
⚡	WallTime	test_benchmarks_lanczos_interp[xval-conserve_dc-run]	140.4 µs	79.3 µs	+77.16%
⚡	WallTime	test_benchmarks_lanczos_interp[kval-conserve_dc-run]	66.1 µs	40.4 µs	+63.47%
⚡	WallTime	test_benchmark_spergel_calcfluxrad[run]	122 µs	78.6 µs	+55.12%
⚡	WallTime	test_benchmark_invert_ab_noraise[run]	133.9 µs	86.7 µs	+54.39%
⚡	WallTime	test_benchmark_spergel_init[run]	41.3 µs	27.3 µs	+50.99%
⚡	WallTime	test_benchmark_rng_discard[run]	121.3 µs	85.2 µs	+42.27%
⚡	WallTime	test_benchmarks_lanczos_interp[kval-no_conserve_dc-run]	57.1 µs	41.1 µs	+39.1%
⚡	WallTime	test_benchmark_gaussian_init[run]	7.3 µs	5.3 µs	+37.84%
⚡	WallTime	test_benchmark_spergel_init_float[run]	7.3 µs	5.4 µs	+36.43%
⚡	WallTime	test_benchmark_interpimage_flux_frac[run]	439.4 µs	332.2 µs	+32.28%

Tip

Curious why performance improved? Comment @codspeedbot explain why performance improved on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing AdamField118:sinc-kval-stencil (a04b2de) with main (e44e1a9)}