Keep integer dtype in mercator_transform

[ybai08]

Problem

mercator_transform always returns a float64 array. When a uint8 image is passed to ImageOverlay(..., mercator_project=True), write_png then sees float data and rescales every channel to its own maximum, so the colours change. A uniform (100, 50, 200, 255) image comes out white:

import numpy as np
import folium

data = np.full((4, 3, 4), (100, 50, 200, 255), dtype="uint8")
bounds = [[40, 0], [50, 10]]
plain = folium.raster_layers.ImageOverlay(data, bounds)
projected = folium.raster_layers.ImageOverlay(data, bounds, mercator_project=True)
# plain pixel:     (100, 50, 200, 255)
# projected pixel: (255, 255, 255, 255)

Changes

If the input has an integer dtype, round the interpolated result and cast it back to that dtype. Float input is unchanged, so the existing test_image_overlay URL stays the same.

Two regression tests were added: mercator_transform keeps uint8 data unchanged for a uniform image, and ImageOverlay produces the same URL for that image with and without mercator_project.

Validation

• Without the fix (folium/utilities.py reverted), both new tests fail: assert dtype('float64') == <class 'numpy.uint8'> and a URL mismatch. With the fix both pass.
• python -m pytest tests --ignore=tests/selenium --ignore=tests/snapshots: 397 passed, 3 failed. The 3 failures are in tests/test_repr.py (cannot import name 'webdriver' from 'selenium', since selenium isn't installed) and fail the same way on main.
• ruff check folium tests (0.16.10), black --check folium tests (26.10.0), codespell on the changed files, and mypy folium pass.
• The selenium, snapshot, and notebook suites were not run.