intel/fs/xe2+: Don't lower barycentric load offsets to fixed-point format on Xe2+.

Floating-point offsets work fine in combination with the
floating-point arithmetic we're about to lower these intrinsics into,
and they require less instructions than converting to fixed-point and
then back.  No reason to take the precision/range hit nor the extra
instructions.

Reviewed-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29847>

src/intel/compiler/brw_nir.c

@@ -618,9 +618,10 @@ brw_nir_lower_fs_inputs(nir_shader *nir,
                                    NULL);
    }
 
-   nir_shader_intrinsics_pass(nir, lower_barycentric_at_offset,
-                                nir_metadata_control_flow,
-                                NULL);
+   if (devinfo->ver < 20)
+      nir_shader_intrinsics_pass(nir, lower_barycentric_at_offset,
+                                 nir_metadata_control_flow,
+                                 NULL);
 
    /* This pass needs actual constants */
    nir_opt_constant_folding(nir);