DiffPrivacy ICL Pack - OpenAI Completion LLMs (#11881)

* add logprob

* version bump

* typo

* rename pack

* add first implementation of dp algo - gaussian

* fix minor errors

* modify templates

* tokenizer

* add logprobs for completions api

* way better results with completions

* add more dynamic few shot examples

* incorporate dynamic few shot

* remove chat completion stuff

* add laplacian mechanism

* add latest core and llms-openai

* demo nb -f

* conflict

* pants

* pyproject metadata

* rename pack

* get rid of old pack

* fix tqdm

* dispatcher

* add events

* improved async

* add start/end events

* add sizes int option

* add example gen script for symptom_2_disease

* add final save

* pants

* remove unused span handler, and make event handler log error report every error

* add num synthetic starts to error report

* report the EventHandler itself

* increase threshold to ~2.5% error rate

* change to ratio of errors

* actually count event ends

* update main

* change number of size to 3 instead of 4

* add retriever demo; add node prost processor

* pants

* add sigma_to_eps

* remove networks stuff

* add back basic demo

* add writeups for basic demo

* rm data

* rm demo usage from ignored dir /notebooks

* cleaning symptom_2_disease

* Add README and sample synthetic_dataset.json

* rm lock file

* remove extra error_report.json

* add README details

* remove empty utils file

* use exponential mechanism instead of laplace

* remove data files add download script

* remove big json files

* update privacy mechanism

* fix sigma_to_eps

* remove synthetic_dataset.json

* improve dp description

* update authors

* update

* minor fix

* minor fix

* minor validation