CI: Use windows-2025 in the GMT Dev Tests workflow

[seisman]

Description of proposed changes

xref: https://github.blog/changelog/2024-12-19-windows-server-2025-is-now-in-public-preview/

[weiji14]

It's not even 2025 yet 😆

[seisman]

There is one failure in the GMT Dev Tests workflow, but it's unrelated to changes in this PR.

=================================== FAILURES ===================================
________________________ test_sphinterpolate_no_outgrid ________________________

mars =      longitude  latitude  radius_m
0      278.231   -40.059      6888
1      302.360    15.487      -233
2       57.10...8.384   -32.260      1375
368    312.032    22.271     -4400
369    134.260    47.670     -3820

[370 rows x 3 columns]

    @pytest.mark.benchmark
    def test_sphinterpolate_no_outgrid(mars):
        """
        Test sphinterpolate with no set outgrid.
        """
        temp_grid = sphinterpolate(data=mars, spacing=1, region="g")
        assert temp_grid.dims == ("lat", "lon")
        assert temp_grid.gmt.gtype == 1  # Geographic grid
        assert temp_grid.gmt.registration == 0  # Gridline registration
        npt.assert_allclose(temp_grid.max(), 14628.144)
        npt.assert_allclose(temp_grid.min(), -6908.1987)
        npt.assert_allclose(temp_grid.median(), 118.96849)
>       npt.assert_allclose(temp_grid.mean(), 272.60593)
E       AssertionError: 
E       Not equal to tolerance rtol=1e-07, atol=0
E       
E       Mismatched elements: 1 / 1 (100%)
E       Max absolute difference among violations: 0.00024763
E       Max relative difference among violations: 9.0836[965](https://github.com/GenericMappingTools/pygmt/actions/runs/12424532581/job/34689838711?pr=3701#step:15:966)e-07
E        ACTUAL: array(272.60568, dtype=float32)
E        DESIRED: array(272.60593)