Don't create copy of variable when reverse indexing with indexing overloaded

[Book-reader]

fixes #3258.

Please let me know if there is a better way to do this.

This is what the LLVM IR generated from https://github.com/Book-reader/c3c/blob/aeeb35c46702c79768cc4e8f6332769949a1874f/test/test_suite/overloading/reverse_index.c3t looked like before these changes:

define void @test.main() #0 {
entry:
  %t = alloca %Test, align 8
  %.anon = alloca %Test, align 8
  %.anon2 = alloca %Test, align 8
  call void @llvm.memcpy.p0.p0.i32(ptr align 8 %t, ptr align 8 @.__const, i32 16, i1 false)
  call void @llvm.memcpy.p0.p0.i32(ptr align 8 %.anon, ptr align 8 %t, i32 16, i1 false)
  %0 = call i64 @test.Test.len(ptr %.anon)
  %sub = sub i64 %0, 1
  %1 = call i32 @test.Test.get(ptr %.anon, i64 %sub)
  %2 = call i64 @test.Test.len(ptr %t)
  %sub1 = sub i64 %2, 1
  %3 = call i32 @test.Test.get(ptr %t, i64 %sub1)
  call void @llvm.memcpy.p0.p0.i32(ptr align 8 %.anon2, ptr align 8 %t, i32 16, i1 false)
  %4 = call i64 @test.Test.len(ptr %.anon2)
  %sub3 = sub i64 %4, 1
  %5 = call ptr @test.Test.get_ref(ptr %.anon2, i64 %sub3)
  %6 = call i64 @test.Test.len(ptr %t)
  %sub4 = sub i64 %6, 1
  %7 = call ptr @test.Test.get_ref(ptr %t, i64 %sub4)
  %8 = call i32 (ptr, ...) @printf(ptr @.str, i32 %1, i32 %3, ptr %5, ptr %7)
  ret void
}

It would allocate a new copy of the original value named %.anon* for each reverse-indexing, making &t[^1] return a pointer to a different value each time it was used.

This doesn't affect Lists in most cases because they store their contents on the heap and the pointer to it would be shared between the copies, but appending to a List using arr[^0] = x would not update the length of the original list and could reallocate the array and cause the original List to point to freed memory.

[lerno]

This does not quite solve toe problem. What you need to do is get a reference (if possible) to it and then use that instead. I've pushed another fix.