Convert all consecutive calls to d7_DrawPrimitive(TRIANGLE_FAN) into
a single call to d7_DrawPrimitive(TRIANGLE_LIST) with all the vertices.
Note, it *increase* the number of vertices, but bandwith is much less costly
than multiple calls.
Note, only a very precise subset of the calls get buffered in order to
ensure that the disruption is minimal.
Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=33814
--
v19: ddraw: Increasing the vertex batch size on demand
ddraw: add d3d_perf statistics on buffering
wined3d: Add a TRACE in wined3d_streaming_buffer_unmap()
ddraw: Add a local buffer in d3d_device7_DrawPrimitive()
https://gitlab.winehq.org/wine/wine/-/merge_requests/2105