[BugFix] Robustize cuFINUFFT Python guru interface.

[SepandKashani]

As per the docs:

• finufft.Plan(n_modes_or_dim) accepts (int, tuple[int])
• cufinufft.Plan(n_modes) accepts (int, tuple[int])

In practice n_modes_or_dim is transformed internally to a NumPy array, hence providing any iterable to finufft.Plan() works.
However n_modes in cufinufft.Plan() is more restrictive: it works with tuple[int], fails planning if list[int], and wrongfully passes planning if ndarray[int], but then seg-faults at execute().

The examples below showcase the problem:

### mve_finufft.py
import finufft
import numpy as np

xp, fi = np, finufft

M, D = 1_000, 2
N = (20, 21, 22)[:D]
cdtype = np.cdouble

plan = fi.Plan(
    nufft_type=1,
    n_modes_or_dim=tuple(N),  # ok
    # n_modes_or_dim=list(N),  # ok
    # n_modes_or_dim=np.array(N),  # ok
    dtype=cdtype,
    isign=1,
)

rng = xp.random.default_rng(0)
x = rng.uniform(-np.pi, np.pi, size=(D, M))
plan.setpts(**dict(zip("xyz"[:D], x[:D])))

w = rng.standard_normal(2 * M).view(cdtype)
out = plan.execute(w)

### mve_cufinufft.py
import cufinufft
import cupy as cp
import numpy as np

xp, fi = cp, cufinufft

M, D = 1_000, 2
N = (20, 21, 22)[:D]
cdtype = np.cdouble

plan = fi.Plan(
    nufft_type=1,
    n_modes=tuple(N),  # ok
    # n_modes=list(N),  # planning error
    # n_modes=np.array(N),  # execute error + seg-fault (sometimes)
    dtype=cdtype,
    isign=1,
)

rng = xp.random.default_rng(0)
x = rng.uniform(-np.pi, np.pi, size=(D, M))
plan.setpts(**dict(zip("xyz"[:D], x[:D])))

w = rng.standard_normal(2 * M).view(cdtype)
out = plan.execute(w)

This PR increases the robustness of cufinufft.Plan() to match the flexibility of the CPU interface.

[janden]

Looks great. Thanks!