intel/compiler: insert URB fence before task/mesh termination

Bspec 53421 says:
"A URB fence memory is typically performed prior the thread
exit message, so that the next thread dispatch that reads
that URB memory will see it."

Cc: 22.1 <mesa-stable>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16665>

src/intel/compiler/brw_fs.h

@@ -327,6 +327,7 @@ public:
    void emit_gs_input_load(const fs_reg &dst, const nir_src &vertex_src,
                            unsigned base_offset, const nir_src &offset_src,
                            unsigned num_components, unsigned first_component);
+   void emit_urb_fence();
    void emit_cs_terminate();
    fs_reg emit_work_group_id_setup();