feat: add optional non-breaking ViT-L/14@336px support

[yurijmikhalevich]

For those who wants extra search quality and ready to sacrifice speed for it:

Stats from a dataset of 269 photos on a much slower NAS (using the v1.2.3 and v2.0.0-pre AppImage distributions):

• ingestion: 3m56.626s vs 125m0.507s (+3070%)
• search: 0m18.064s vs 3m19.742s (+1006%)

More details are here.

Like and comment if you want to see the largest CLIP model in rclip.

[unanthropomorph]

I am a bit late to this but since you're already using openclip I'd recommend using laion/CLIP-ViT-B-16-DataComp.XL-s13B-b90K or laion/CLIP-ViT-B-32-256x256-DataComp-s34B-b86K as they yield only a slightly lower performance than ViT-L/14@336 with roughly 1/3rd of the parameters.
Even better, give the user the option to pick an openclip model from huggingface.

[yurijmikhalevich]

@unanthropomorph, thank you for flagging this! Yes, this is what I was considering, too.