rasterize: resolution= argument not honored when bounds don't divide evenly

[brendancol]

Describe the bug

When resolution= is passed to rasterize(), width and height come from ceil((xmax - xmin) / x_res), but output coordinates are then rebuilt over the original bounds. If the bounds don't divide evenly by the resolution, the actual cell size ends up smaller than what was asked for.

Example:

from shapely.geometry import box
from xrspatial.rasterize import rasterize

r = rasterize([(box(0,0,1,1), 1.0)], resolution=0.3, bounds=(0,0,1,1), fill=0)
# shape (4, 4); actual cell size = (1 - 0) / 4 = 0.25, not 0.3
print(r.x.values[1] - r.x.values[0])  # 0.25

Caller asked for 0.3-unit pixels, got 0.25-unit pixels.

Expected behavior

Either honor resolution= exactly (expand the output bounds so each cell is genuinely 0.3 units; matches rasterio / GDAL from_bounds / from_origin semantics), or document the parameter as a maximum cell size with a note that the actual size may shrink to fit the input bounds.

Code location

• xrspatial/rasterize.py:3223 -- width/height computed with ceil(bounds / resolution)
• xrspatial/rasterize.py:3352 -- output coords rebuilt over original xmin..xmax

Proposed fix

Take the first path. When resolution= drives width/height, set xmax_out = xmin + final_width * x_res and ymax_out = ymin + final_height * y_res before building coords. Existing tests with bounds that divide evenly pass unchanged.