OptimalSteps scheduler improvements

[EricBCoding]

Hi,

I introduced a few quality of life improvements to the OptimalSteps scheduler node from #7584 :

• Reduced the minimum value of steps from 3 to 1: there are times when you may want to run inference with only 1-2 steps, such as 2-step upscaling @ 50% denoise.
• Added an SDXL preset to the NOISE_LEVELS dictionary: the values here are derived from the AYS scheduler and seem to work pretty well.
• Added custom_sigmas optional input: the log-linear interpolation included with this node may prove useful for other models, especially if you can pass in your own sigma values. Supplying a comma-separated string will override the included presets (SDXL, WAN, Flux).

[henrikvilhelmberglund]

Is 0.0 correct? This gives a black image for me, changing it to 0.001 like the others fixed it but not sure what the correct value should be.

[EricBCoding]

Oh, good call. It should probably end with 0.001 for better compatibility across samplers. Thank you, @henrikvilhelmberglund !