Closes #300

Problem

core._adjust_for_anisotropy subtracts the (float) center from the input coordinates in place:

X -= center

When the coordinates are integers — e.g. OrdinaryKriging(x, y, z) with integer x/y, or integer query points passed to execute("points", ...) — NumPy (≥ 2) refuses the in-place subtraction under the same_kind casting rule:

numpy._core._exceptions._UFuncOutputCastingError: Cannot cast ufunc 'subtract'
output from dtype('float64') to dtype('int64') with casting rule 'same_kind'

Reproduction (current main, NumPy 2.4)

import numpy as np
from pykrige.ok import OrdinaryKriging

x = np.array([10, 20, 30, 40, 50])   # integer coordinates
y = np.array([10, 25, 35, 45, 60])
z = np.array([1.1, 2.2, 0.5, 3.1, 1.7])
OK = OrdinaryKriging(x, y, z, variogram_model="linear")
OK.execute("points", np.array([15, 25]), np.array([15, 25]))   # -> UFuncTypeError

Fix

Cast X to float64 at the start of _adjust_for_anisotropy. The docstring already documents X as a float array, so this simply enforces that contract. All Ordinary/Universal Kriging classes (2D and 3D) route their coordinate handling through this single function, so the one-line fix covers every variant.

Verification

• Integer-coordinate kriging now returns results identical to the equivalent float-coordinate run (2D and 3D), confirmed numerically.
• Added test_core_adjust_for_anisotropy_integer_coords (unit-level + an end-to-end OrdinaryKriging run); it fails on main with the UFuncTypeError and passes with this change.
• Full tests/ suite: 54 passed, 2 skipped (no new failures). black / isort clean.