Hook up new LLM abstraction to LLMPredictor (#6685)

* wip

* wip

* wip

* wip

* wip

* add notebook

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wi

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* rip out token counter

* wip

* wip

* wip

* remove formatted prompt logging

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wipg

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* better typing

* fix bug

* wip

* wip

* wip

* simplify and rename

* wip

* wip

* wip

* wup

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* allow tiktoken for python38

* wip

* wip

* address comments

* fix lint

* add note

* fix merge

* wip

* wip

* wiup

* add to agent and programs

* wip

* update notebook

* wip

* wip

* wiup

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* remove unused imports

* update change log

* add back missing file

* clean

* wip

* update notebook

* update notebook

* wip

* expose message history

* wip

* wip

* revert chat engine changes

* wip

* lint