The IStorage (DocFile) implementation was adding empty blocks at the
end of the file every time a caller wrote data that was not a multiple
of 512 bytes.  If the caller made a lot of very small writes the file
size could be huge even though the data in the file was tiny. This was
caused by BlockChainStream_SetSize trying to allocate file blocks for
the new data using a condition that bore no relationship to the
required condition, and it was not necessary to do so at that time
since it is done (the right way) by StorageImpl_GetNextFreeBigBlock
(called via BlockChainStream_Enlarge).

Windows doesn't like when a thread plays games with the focus, that
leads to all kinds of misbehaviours and failures to activate a
window. So, better don't generate a mouse click message in
mouse_ll_global_thread_proc.

Implement some basic opengl accelerated blts from and to render
targets. It's not perfect yet, but enought to make some D3D apps
happy. For now the only supported operations are:

- Full screen back -> Front buffer: Just call present
- Offscreen surface -> render target
- Render target -> offscreen surface(slow)
- render target colorfill

Release the outer unknown in the proxybuffer test to show that the
last release of the outer unknown doesn't free the proxy buffer.

Release the proxy buffer object on the last release of the proxy's outer
unknown, otherwise memory is leaked for the Ndr implementation of
IRpcProxyBuffer.

Fix up the hand-coded proxies to match the behaviour from the Ndr
implementation.

rpcrt4: Fix copying too much data from the buffer by not taking into account the fact that auth_pad_len has been included in the fragment length.

Protect LPGUID typedef by __LPGUID_DEFINED__ being defined.

Protect LPCLSID and LPFMTID typedefs, IsEqualGUID, IsEqualIID,
IsEqualCLSID, IID_NULL, CLSID_NULL and FMTID_NULL defines by
__IID_DEFINED__ being defined.