#4448 fix checkpoint cache usage

Add username and password in ngrok

4415 update generation info

Adding optional tag shuffling for training

Add DeepDanbooru to the interrogate API

- if checkpoint cache is set to 1, keep 2 models in cache (current +1 more)

- cache model after is has been loaded from file

Where are they now? Here: https://github.com/AUTOMATIC1111/stable-diffusion-webui-old-localizations

Add a barebones CLIP interrogate API endpoint

list changed Settings keys in UI
do not print VRAM/etc stats everywhere but in calls that use GPU

Fixes #800 #1562 #2075 #2304 #2931 LDSR upscaler producing black bars

Italian localization - Additions and updates

Updated localization with the latest version of these Scripts/Extensions:

SD-Chad - Stable Diffusion Aesthetic Scorer (added)
AlphaCanvas
StylePile
Alternate Sampler Noise Schedules
SD Latent Mirroring (new)
SD Dynamic Prompts
Dataset Tag Editor
Depth Maps (new)
Improved prompt matrix (new)

New Extensions:

Auto-sd-paint-ext (new)
training-picker (new)
Unprompted
NovelAI-2-local-prompt (new)
tokenizer (new)

Hope there aren't too many mistakes or wrong translations, in case let me know.

Update ko_KR.json

New setting option and some additional extension index strings

Add support for the new DPM-Solver++ samplers added to k-diffusion

Python 3.8 typing compatibility