Type unification merges aliases for generic types with pointers for cpp backend

[zacharycarter]

Let's say you have some C++ code like:

typedef struct AliasOne_T *AliasOne;
typedef struct AliasTwo_T *AliasTwo;

std::vector<AliasOne> foo() {
  std::vector<AliasOne> f(1);
  return f;
};

std::vector<AliasTwo> bar() {
  std::vector<AliasTwo> b(1);
  return b;
};

And you want to wrap it with Nim.

You'd define some type definitions and proc signatures like:

type
  AliasOne {.importcpp.} = pointer
  AliasTwo {.importcpp.} = pointer

proc foo(): StdVector[AliasOne] {.importcpp.}
proc bar(): StdVector[AliasTwo] {.importcpp.}

What happens during compilation is the two alias type gets merged since they're both just a pointer to an opaque type, and you end up with only the type for the first used symbol in the generated C++ code.

So instead of seeing:

typedef std::vector<VkImage> TY__SomeRandomSuffix;
typedef std::vector<VkImageView> TY__SomeOtherRandomSuffix;

in the generated cpp source, you will only see something like:

typedef std::vector<VkImage> TY__SomeRandomSuffix;

And if you reference VkImageView first instead of VkImage you'd see the reverse - the alias for the VkImageView vector would be present but not the one for the VkImage vector.

This causes the generated C++ to fail to compile, and means one has to resort to hacks like {.emit.} which produces other side effects - like now there's the need to define the std::vector interface for each aliased type in Nim.

Issue is blocking my Vulkan renderer / game which you can see some of the WIP code for here: https://github.com/zacharycarter/FRAG/blob/master/src/fragpkg/app.nim#L52-L56 - in fact this is the section of code I'm blocked from progressing further on (barring resorting to emit).

I tested this and reproduced it on 1.4.8 and 1.5.1

Nim Compiler Version 1.5.1 [Windows: amd64]
Compiled at 2021-09-21
Copyright (c) 2006-2021 by Andreas Rumpf

git hash: 928ea6bb4c0cc791f1b81e55dbebd14d5c6a7315
active boot switches: -d:release

[metagn]

Related? #11797

[metagn]

Since #24313 distinct should maybe be usable here