Add atomic point quality dataset generator

[comcon1]

This script generates the point quality dataset of the following schema:

c	h	val_s	err_s	val_e	err_e
Carbon SMILE ID	H ID	OP value (simulatioon)	err	OP value (experiment), with the sign from the simulation	err

For a single carbon atom, if proton's OP value cannot be different (there is only one proton or 3 protons), then H ID is only "1" and only single value is used.

For CH2 groups, protons can be different, we assign 1 to be proton with smaller absolute OP and 2 with larger.
This clears the problem of the uncertainty of chirality that we have for some atoms.

We do not trust the sign of the OP value in any experiments. Signs are always inherited from the simulation.

All values from all experiments are averaged using only c,h pair as a key.

---

📚 Documentation preview 📚: https://databank--495.org.readthedocs.build/

[batukav]

Left a few inline comments. The most important one is the hasattr typo in add_experiment_data that resets _exp_opdicts on every call — as written, lipids with multiple paired experiments will only have the last one survive into the average. The rest are smaller cleanups.