[accuracy] AI tagging accuracy improvements (more training, negative/positive reinforcement, audit ...)

Actions requests from the admin to improve the accuracy

• Guidance to write good tag description

• Suggest negative reinforcement for AI tagging (either raising it to the admin or training the AI)

• Let user flag good examples of conversations that should have the tag

• Allow to train on more than 10 tags

• How to audit/re-validate samples for a given tag

Core accuracy

• ignore signature

• ignore email inside emails or forwarded emails