r3.gwflow: fix segmentation fault

[jayneel-shah18]

This PR resolves a segmentation fault in the r3.gwflow module that occurs when the optional sink parameter is not provided. The issue was caused by unconditional dereferencing of param.q->answer and use of data->q, even when the q array was never allocated.

Fix Summary

• Memory Allocation:

  • Changed:

    N_alloc_gwflow_data3d(cols, rows, depths, 0, 0);

    to:

    N_alloc_gwflow_data3d(cols, rows, depths, 1, 0);

    to ensure data->q is always allocated.

• Runtime Check:

  • Replaced the sink reading line:

    N_read_rast3d_to_array_3d(param.q->answer, data->q, param.mask->answer);

    with:

    if (param.q->answer != NULL) {
        N_read_rast3d_to_array_3d(param.q->answer, data->q, param.mask->answer);
        N_convert_array_3d_null_to_zero(data->q);
    } else {
        G_message(_("No sink map provided. Initializing zero sink..."));
        for (z = 0; z < geom->depths; z++) {
            for (y = 0; y < geom->rows; y++) {
                for (x = 0; x < geom->cols; x++) {
                    N_put_array_3d_d_value(data->q, x, y, z, 0.0);
                }
            }
        }
    }

r3.gwflow now works as expected without sink parameter. This PR fixes #6142.

[petrasovaa]

Why the need for the change to the N_alloc_gwflow_data3d call?

[jayneel-shah18]

The change ensures data->q is safely allocated when needed and aligns with how other optional conditions are handled. It also prevents undefined behavior and keeps the memory layout consistent for future extensions.