Handler-less kernel submit path (parallel_for with nd_range)

[slawekptak]

This PR introduces a fully handler-less kernel submission path. The feature is not complete yet. For testing purposes we introduce the __DPCPP_ENABLE_UNFINISHED_NO_CGH_SUBMIT macros to enable unit tests for the new handler-less path. This macro should not be used by the application, and a legacy handler-based path is used. Once the handler-less path is fully implemented, we will switch corresponding APIs to use it unconditionally and will remove the macros.

This PR covers:

1. A parallel_for, nd_range based kernel submit.
2. Parallel_for queue shortcut, enqueue_functions extension and KHR free functions extension.
3. A scheduler-based kernel submission.
4. A new unit test which covers the host task and kernel ordering for an in-order queue (including the handler-less path).

[vinser52]

In this PR, I would like to see at least one public interface implementation that utilizes this approach, just to ensure it works.

[slawekptak]

In this PR, I would like to see at least one public interface implementation that utilizes this approach, just to ensure it works.

In the latest update, there are two public interfaces: The enqueue functions extension, and queue.parallel_for. Both are enabled only if __DPCPP_ENABLE_UNFINISHED_NO_CGH_SUBMIT is defined.

[uditagarwal97]

LGTM.

[github-actions]

@intel/llvm-gatekeepers please consider merging

[Alexandr-Konovalov]

Could the text be altered in the subsequent patches, as this code can be called not only from parallel_for?

[slawekptak]

Yes, the plan is to extend this to other functions once parallel_for(nd_range) is complete.