Query modules in C++ are being congested by the unique lock

[Josipmrden]

Memgraph version
v2.16.0

Environment
Ubuntu 22.04

Describe the bug
It seems that the memory dispatch guard is taking around 30% time on holding the unique lock while registering and unregistering the thread. Furthermore, the Register and Unregister methods are both called as many times as the query module is run (so at least once per query, but can be more)

Query module for testing

#include <mgp.hpp>
#include <string>

void isEqual(mgp_list *args, mgp_func_context *ctx, mgp_func_result *res, mgp_memory *memory) {
    mgp::MemoryDispatcherGuard guard(memory);

    const auto arguments = mgp::List(args);
    auto result = mgp::Result(res);

    std::string output = std::string(arguments[0].ValueString());
    std::string input = std::string(arguments[1].ValueString());

    result.SetValue(output == input);
}
 
extern "C" int mgp_init_module(struct mgp_module *module, struct mgp_memory *memory) {
  try {
    mgp::MemoryDispatcherGuard guard(memory);
 
    mgp::AddFunction(isEqual, "is_equal",
                     {mgp::Parameter("output", mgp::Type::String), mgp::Parameter("input", mgp::Type::String)}, module, memory);
  } catch (const std::exception &e) {
    return 1;
  }
  return 0;
}
 
extern "C" int mgp_shutdown_module() {
  return 0;
}

To Reproduce
Steps to reproduce the behavior:

1. foreach (i in range(1, 10000000) | CREATE (:Node {id1: toString(i), id2: toString(i)}));
2. Observe how much time it takes to execute match (n) where equality.is_equal(n.id1, n.id2) = true return count(*);
3. observe how much time it takes to execute the command above in parallel (use 2 consoles and run one immediately after the other) -> the time is more than doubled for both queries

Expected behavior
Queries shouldn't be that slow

Flamegraph