Run each scene in its own context

glmark2 has always run all of its scenes on a single context created and made current during initialization. This is not the preferred way to run benchmarks or even directed functional tests as it allows the results to be affected by any issues with internal driver bookkeeping, etc. (of course, it is an interested way to run the tests to detect any resource drain or leakage). We will implement a reset method on the Canvas object to allow a fresh context to be created for each scene. An option will be available to allow for the aforementioned "leak" testing, but the default will be to create a new context for each scene.