Add RWKV strategy to UI and ability to save with user model data

[dblacknc]

Description

When using an RWKV model, the loading strategy must be expressed on the command line, and is often model size-specific. It also can be a relatively complex argument. Add the ability to set this in the UI (i.e. before loading another RWKV model) and to save those parameters with the user model data in models/config-user.yaml.

Additional Context

I use two 12 GB RTX 3060 GPUs. To load a 7B RWKV model, I typically use something like: --rwkv-strategy "cuda:0 fp16 *17 -> cuda:1 fp16"

For squeezing a 14B model into those GPUs, I might use: --rwkv-strategy "cuda:0 fp16 *16 -> cuda:1 fp16 *8 -> cuda:1 fp16i8"

At a minimum, I'd like to be able to save that strategy argument value in the embedded model config file. Definitive examples are here: https://pypi.org/project/rwkv/ BTW the current RWKV page in this project may have some --rwkv-strategy examples that aren't syntactically valid, and could be updated/corrected.