Test: Unorderded Map Access

[ax3l]

Summary

Seeking help for the issue in #2367 by providing these unit tests as part of check-enzyme-integration-reverse.

This test demonstrates two issues with std::unordered_map:

• operator[] has a memory leak/double free
• at() fails to compile

Compile Warnings & Error

freeing without malloc   %46 = load ptr, ptr %0, align 8, !tbaa !23
freeing without malloc   %46 = load ptr, ptr %0, align 8, !tbaa !23
freeing without malloc   %15 = load ptr, ptr %malloccall, align 8, !tbaa !24
freeing without malloc   %28 = load ptr, ptr %malloccall1, align 8, !tbaa !24
freeing without malloc   %15 = load ptr, ptr %malloccall, align 8, !tbaa !24
freeing without malloc   %28 = load ptr, ptr %malloccall1, align 8, !tbaa !24
freeing without malloc   %11 = load ptr, ptr %3, align 8, !tbaa !16
freeing without malloc   %25 = load ptr, ptr %24, align 8, !tbaa !16
freeing without malloc   %22 = phi ptr [ %23, %33 ], [ %19, %.preheader ]
freeing without malloc   %40 = load ptr, ptr %2, align 8, !tbaa !26
error: <unknown>:0:0: in function preprocess__Z15compute_umap_atd double (double): Enzyme: Cannot deduce type of copy   call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(7) %6, ptr noundef nonnull align 1 dereferenceable(7) @.str, i64 7, i1 false) #21

To Do

Please feel free to push to this PR as needed 🙏 I will also try to mingle around with it a bit, but am still very early in understanding what's going on :)

• add fix for memory leak/double free in operator[]
• add fix for compile issue with at()