Follow-up to JDK-8352209 and PR discussion https://github.com/openjdk/jfx/pull/1951

While above issue was patched using a null check in BaseContext.flushMask(), it might be just a temporary workaround to the real issue.

It seems like the root cause is Prism's BaseContext class not keeping the texture locked until VertexBuffer is fully flushed. This can potentially cause a situation where BaseContext still holds on to an earlier created mask texture and attempts to update it, but Prism pool already freed it since it was not locked. Investigate and fix.