use explicit staging buffer for constant mem optimization and make synchronization less pessimistic

[streichler]

This pull request does two related things:

1. It introduces an explicit staging buffer in pinned host memory for kernel launch data that's destined for
   device constant memory, allowing a clear separation between the part of the transfer that's synchronous
   (i.e. the transfer from the caller's stack to the staging buffer) and the part that's asynchronous (the transfer
   from the staging buffer to actual device constant memory).
   Note that the staging buffer's property of being per-device (as opposed to per-stream or truly global) is
   a bit awkward for the current Kokkos implementation. I'm tracking the constant buffer using static members
   of the CudaInternal class so that they're properly shared by all Kokkos::Cuda instances (that are right now
   all on the same device), but changes to Kokkos to better support multiple devices will need to move these to
   whatever object tracks per-device information.

2. Cleans up the host/device synchronization to be both correct (i.e. actually guarantee that the copy from
   the caller's stack is complete before returning) and more precise (i.e. waiting only for the most recent grid
   launch to use the constant mem buffer rather than waiting for all previously issued CUDA work, whether or
   not its using the constant mem buffer). A side benefit from a Legion+Kokkos interop perspective is that the
   use of cudaEventSynchronize plays more nicely with Legion task parallelism than cudaDeviceSynchroinze
   does.
   Although it is not attempted in this PR, the use of events for the host/device synchronization permits a further
   performance optimization in which multiple grid launches that collectively fit within the 32KB constant mem
   buffer could be permitted to run concurrently. A new launch would then just wait on exactly the events for the
   grids using constant mem buffer locations that are about to be reused. The current implementation represents
   the degenerate case in which every grid is assumed to need the whole 32KB.

[dalg24-jenkins]

Can one of the admins verify this patch?

[dalg24]

OK to test

[streichler]

Changed three occurrences of 0 to nullptr to address build failures.

[dalg24]

LGTM