[Lang] Adds @.atomic and a few implementations

[dmed256]

Description

Adds @atomic for atomic support

Serial and OpenMP modes should be able to do any @atomic operation.

• Serial: Doesn't need any implementation
• OpenMP: Leverages #pragma omp atomic and #pragma omp critical

For the GPU modes, we have some basic implementations done. Here base types includes int, float, and double. The rest should be coming out soon!

Mode	(int) --	(int) ++	(base types) +=	(base types) -=	Atomic blocks
CUDA	✔️	✔️	✔️	✔️	-
HIP	✔️	✔️	✔️	✔️	-
OpenCL	-	-	-	-	-
Metal	-	-	-	-	-

Examples

Checkout the reduction example (04_reduction) where we showcase using @shared and @atomic for two ways to do reductions.

Basic atomic expressions

@atomic i += 1;
@atomic ++i;

Atomic regions

@atomic {
  i += 1;
  i += 1;
}

[codecov]

Codecov Report

Merging #410 into main will decrease coverage by 0.03%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##             main     #410      +/-   ##
==========================================
- Coverage   76.34%   76.30%   -0.04%     
==========================================
  Files         240      241       +1     
  Lines       18425    18545     +120     
==========================================
+ Hits        14066    14151      +85     
- Misses       4359     4394      +35     

Impacted Files	Coverage Δ
src/lang/expr/expr.cpp	49.49% <0.00%> (-2.09%)	⬇️
src/lang/expr/exprNode.cpp	76.92% <ø> (ø)
src/lang/modes/cuda.cpp	66.34% <15.00%> (-32.10%)	⬇️
src/lang/builtins/attributes/atomic.cpp	96.15% <96.15%> (ø)
src/lang/builtins/attributes/tile.cpp	92.61% <100.00%> (ø)
src/lang/modes/openmp.cpp	89.58% <100.00%> (+8.10%)	⬆️
src/lang/parser.cpp	86.56% <100.00%> (+0.01%)	⬆️
src/lang/specialMacros.cpp	59.48% <0.00%> (ø)
src/tools/sys.cpp	65.31% <0.00%> (+0.16%)	⬆️
... and 2 more