rhi: gl: Unify end/finish behavior

Issue a glFlush in endOffscreenFrame. This matches endFrame which
flushes either implicitly (swapBuffers) or explicitly with SkipPresent
(glFlush). This will then benefit producer-consumer setups where a
context renders into a texture using begin/endOffscreenFrame and then
the texture is consumed in another context.

In addition, make finish() issue a glFinish().

Change-Id: I0a3115255ad2ac82b730e26d1ca7e88377c5a28c
Reviewed-by: Piotr Srebrny <piotr.srebrny@qt.io>
Reviewed-by: Andy Nichols <andy.nichols@qt.io>
(cherry picked from commit cb139dbdb8d7ff1f9e9be24623697035677c9774)
Reviewed-by: Qt Cherry-pick Bot <cherrypick_bot@qt-project.org>

src/gui/rhi/qrhigles2.cpp

@@ -1942,6 +1942,12 @@ QRhi::FrameOpResult QRhiGles2::endOffscreenFrame(QRhi::EndFrameFlags flags)
 
     executeCommandBuffer(&ofr.cbWrapper);
 
+    // Just as endFrame() does a flush when skipping the swapBuffers(), do it
+    // here as well. This has the added benefit of playing nice when rendering
+    // to a texture from a context and then consuming that texture from
+    // another, sharing context.
+    f->glFlush();
+
     return QRhi::FrameOpSuccess;
 }
 
@@ -1963,6 +1969,12 @@ QRhi::FrameOpResult QRhiGles2::finish()
             executeCommandBuffer(&currentSwapChain->cb);
             currentSwapChain->cb.resetCommands();
         }
+        // Do an actual glFinish(). May seem superfluous, but this is what
+        // matches most other backends e.g. Vulkan/Metal that do a heavyweight
+        // wait-for-idle blocking in their finish(). More importantly, this
+        // allows clients simply call finish() in threaded or shared context
+        // situations where one explicitly needs to do a glFlush or Finish.
+        f->glFinish();
     }
     return QRhi::FrameOpSuccess;
 }