[#379][Stream]: Add non-blocking stream creation

[deukhyun-cha]

Description

• Add creating a non-blocking stream based on a stream property value "nonblocking"
• Current implementation applies to cuda and hip modes only
• An example is added to demonstrate the usage

[noelchalmers]

I'm not sure I see the value-add of this.

• This option only affects how streams would or would not overlap with work on the NULL stream. But the NULL stream isn't used for any kernels/async APIs in OCCA. Hence this behavior would only be usable in a user's code that is interop-ing OCCA with CUDA or HIP (modulo some blocking memcpys which are always avoidable).
• Consequently, a more valuable and flexible contribution would probably be a device::wrapStream method, rather than additional options at stream creation.
• That new method would be able to support OpenCL's commandQueue options as well.

Just my $0.02 CND.

[tcew]

Is that like $0.01 US?

[amikstcyr]

Thank you for the feedback!

In our opinion, this PR seems:

1. to have roughly zero overhead cost,
2. to have no effect if the underlying API/lib is not syncing with some sort of base stream,
3. to be useful when interfacing with certain libraries,
4. to be ready & responds to a need/use case hence immediately useful.

We could work on your suggestion but not sure why doing so would block this particular PR.

Tim: it's more like 0.015 USD... :-)

[codecov]

Codecov Report

Merging #498 (cea4cba) into main (5db15e0) will increase coverage by 0.05%.
The diff coverage is n/a.

❗ Current head cea4cba differs from pull request most recent head 558be5e. Consider uploading reports for the commit 558be5e to get more accurate results

@@            Coverage Diff             @@
##             main     #498      +/-   ##
==========================================
+ Coverage   76.41%   76.46%   +0.05%     
==========================================
  Files         261      257       -4     
  Lines       19500    19352     -148     
==========================================
- Hits        14900    14797     -103     
+ Misses       4600     4555      -45     

Impacted Files	Coverage Δ
src/c/kernel_fortran_interface.cpp	0.00% <0.00%> (-10.00%)	⬇️
src/occa/internal/utils/store.hpp	90.62% <0.00%> (-6.25%)	⬇️
src/occa/internal/utils/sys.cpp	72.89% <0.00%> (-3.62%)	⬇️
src/dtype/builtins.cpp	44.44% <0.00%> (-2.78%)	⬇️
src/c/base.cpp	84.88% <0.00%> (-2.33%)	⬇️
src/occa/internal/modes/serial/registration.cpp	91.11% <0.00%> (-2.23%)	⬇️
src/occa/internal/utils/env.cpp	81.25% <0.00%> (-1.25%)	⬇️
...a/internal/lang/statement/declarationStatement.cpp	91.91% <0.00%> (-1.09%)	⬇️
src/occa/internal/utils/string.cpp	84.44% <0.00%> (-0.89%)	⬇️
src/occa/internal/lang/specialMacros.cpp	60.26% <0.00%> (-0.67%)	⬇️
... and 9 more