Training fixes for MPS

Improve InvokeAI cross attention reliability and speed when using MPS for large images

fix hints file typo

fix: xformers

Check if q.shape[0] * q.shape[1] is 2**18 or larger and use the lower memory usage MPS optimization if it is. This should prevent most crashes that were occurring at certain resolutions (e.g. 1024x1024, 2048x512, 512x2048).

Also included is a change to check slice_size and prevent it from being divisible by 4096 which also results in a crash. Otherwise a crash can occur at 1024x512 or 512x1024 resolution.

When saving training results with torch.save(), an exception is thrown:
"RuntimeError: Can't call numpy() on Tensor that requires grad. Use tensor.detach().numpy() instead."

So for MPS, check if Tensor.requires_grad and detach() if necessary.

LDSR improvements - cache / optimization / opt_channelslast

Fix token counter color on dark theme

Fix WebUI not working inside of iframes

Support changing checkpoint and vae through override_settings

Add Clear Prompt button to roll_col 

Update SD Upscaler to include user selectable Scale Factor

Atomically rename saved image to avoid race condition with other processes

Update launch.py

allow randomized seeds in prompt_matrix

Add option to avoid sending size between interfaces.

add gradio queueing on ui by default to avoid timeout on client side when share=True

Fix None option of VAE selector

add noise strength parameter similar to NAI

Merger ram usage

Add bracket checking functionality to the txt2img/img2img prompts

macOS install improvements

Reinstate DDPM V1 to LDSR