A few tweaks to reduce peak memory usage, the biggest being that if we
aren't using the checkpoint cache, we shouldn't duplicate the model
state dict just to immediately throw it away.

On my machine with 16GB of RAM, this change means I can typically change
models, whereas before it would typically OOM.

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

- fix according to https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/3348#issuecomment-1287923525

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

Co-authored-by: liggest <43201720+liggest@users.noreply.github.com>

- update to latest

- unify the translation of upscale / upscaler
- unify the translation of seed / random seed
- correct images embedding's translation

- fix translation of Tile / Tiling
- fix translation of Negative prompt

in the bug report form

Update KR localization

it's good enough to merge 🎉

translated with deepl, it needs some tweaks I will fix 🔧

add script callback for before image save and change callback for after image save to use a class with parameters