Version 1.0.8 release

Release v1.0.7

- stream: Disable variable grain size for Tapir in cilk_for benchmarks
- stream: Remove unnecessary noinline statements
- pointer_chase: Improved initialization
- New benchmarks:
    spawn_rate: measures function spawn rate
    hot_range: Like GUPS, except targets a small array
- Scripts updated to python3

Bug fixes for CilkPlus toolchain

Release v1.0.5 was a temporary hack that branched off of master.
v1.0.4 is the true ancestor to this release.

Changes relative to v1.0.4:
- Added dummy macro for cilk_spawn_at so we can still compile with the old
toolchain.
- Use mw_localmalloc instead of malloc (which is size-limited)
 to create the block shuffle index arrays.
- Use hand-written loop to scatter index arrays instead of
 emu_replicated_array_init.
- Use cilk_spawn_at in the right places. Previous release was giving multiple
 threads the same portion of the list.
- Improved logging during validation
- Don't rely on va_arg to get a pointer to the global data structure inside
relink_worker_1d. This is a workaround for emusolutions/sc-driver-tests#342.

Upgraded pointer_chase to work with the CilkPlus toolchain

Contains some hacks and workarounds that should be fixed soon,
the pointer_chase link step is done serially now.

Use cilk_spawn_at for remote spawns (requires CilkPlus toolchain)

Determines LLVM_CILK via CMake instead of environment

Updated for v18.09 toolchain release

Changed CMake required version

Initial release