Extended Profiling API

[vchuravy]

The profiling APIs have undergone some evolution: https://docs.nvidia.com/cupti/main/main.html#evolution-of-the-profiling-apis

In #1826 and #1823 as well as https://github.com/vchuravy/NVPerfWorks.jl I tried targetting the PerfWorks API, which now seems to be "deprecated"

Instead one ought to expose the CUPTI Range Profiling API which looks reasonable enough https://docs.nvidia.com/cupti/main/main.html#range-profiling-api also see examples/CUPTI/samples/range_profiling

The other API to take a look at is the CUPTI Profiler Host API: https://docs.nvidia.com/cupti/main/main.html#cupti-profiler-host-api

cc: @gbaraldi @giordano