Performance ideas

Ideas:
- [x] indicates dead end/implemented
---

- [x] Consider using hashbrown instead of HashMap
- [x] lookup table
- [ ] faster graph_isomorphism checks (modify `petgraph`)
- [x] pgo
- [x] (better) thread pooling (better thread to core matching)
- [x] use with_capacity to reduce allocs
- [x] use reserve to reduce allocs where capacity size is unknown
- [ ] faster MEX algo, possibly using SIMD
- [x] avoid adding edges to nodes that are about to be deleted
- [x] different memory allocator

---

- [ ] prevent unnecessary memory allocation through a non-recursive implementation and mutable structs