intel: Flush when mapping buffer objects so writes don't get reordered.

While GEM covers this for execution it knows about, it doesn't know about
the batchbuffer we're preparing.  Fixes piglit vbo-map-remap.c testcase.

src/mesa/drivers/dri/intel/intel_buffer_objects.c

@@ -243,6 +243,9 @@ intel_bufferobj_map(GLcontext * ctx,
       return obj->Pointer;
    }
 
+   if (!read_only)
+      intelFlush(ctx);
+
    if (intel_obj->region)
       intel_bufferobj_cow(intel, intel_obj);
 

src/mesa/drivers/dri/intel/intel_regions.c

@@ -111,6 +111,8 @@ debug_backtrace(void)
 GLubyte *
 intel_region_map(struct intel_context *intel, struct intel_region *region)
 {
+   intelFlush(&intel->ctx);
+
    _DBG("%s %p\n", __FUNCTION__, region);
    if (!region->map_refcount++) {
       if (region->pbo)