update to the latest vertex fetch proposal

The main difference is that XGL_SLOT_VERTEX_INPUT is replaced by
xglCmdBindVertexData.  This actually simplifies the driver, which can be seen
with:

 $ git show icd/intel/

The new proposal also adds some new formats, but they are ignored for now
because they are marked TBD.
diff --git a/icd/intel/pipeline.h b/icd/intel/pipeline.h
index 3cafae1..7e130ea 100644
--- a/icd/intel/pipeline.h
+++ b/icd/intel/pipeline.h
@@ -72,14 +72,12 @@
     XGL_UINT resource_count;
     XGL_UINT uav_count;
     XGL_UINT sampler_count;
-    XGL_UINT vb_count;
 
     /*
      * rt_count slots +
      * resource_count slots +
      * uav_count slots +
-     * sampler_count slots +
-     * vb_count slots
+     * sampler_count slots
      */
     struct intel_pipeline_rmap_slot *slots;
     XGL_UINT slot_count;