Eloisa's block-averaged topography and some minor fixes

[raluy]

This PR introduces a new option for block-averaged topography and fixes some small issues related to interpolation in SimGrid.py.

Main changes

1. Fix for interpolation domain when interp_flag = 1

The computation of x_e and y_e has been corrected for non-integer resolution ratios (d_xi/dx_inter). Previously, the selected source subdomain could be too small when interpolation was used, leading to insufficient spatial coverage and edge artifacts in the interpolated fields.

2. Fix in building mask interpolation

When interp_flag == 1, the interpolator was using an undefined variable (data_bmask). This has been corrected to use the appropriate source field (bdg_heights).

3. New option: block-averaged topography

This new option has been added via simgrid.json ("topo_average_opt": 1, with "topo_average_opt": 0 by default). When enabled, topography is computed using an area-weighted block average instead of decimation or interpolation. This helps mitigate potential aliasing in terrain data. For interp_flag == 0, uses block_reduce with mean. For interp_flag == 1, uses the function block_average_topo (introduced in couplingUtils.py) based on cell overlap.

Testing

The changes have been tested with the four possible combinations:

• Integer and non-integer resolution ratios (interp_flag = 0 and interp_flag = 1).
• Both interpolation and averaging options (topo_average_opt = 0 and topo_average_opt = 1).

[domingom]

Change 'func=np.mean' by 'func=np.median'