Add a "tips & tricks" page

[leofang]

• many cuda.core operations do not actually need an active CUDA context (aka a device that is set to current)
  • some just need CUDA to be initialized, meaning cuInit(0) is called (ex: querying device properties)
  • some do not even need CUDA initialized or driver/GPU to be present, if the options are passed correctly (ex: JIT compiling via Program)
• cuda.core keeps the context stack size to be either 0 or 1
• ...