Sd-mecha

- fix a bug where the model configurations would silently fail to parse
- fix none seed not supported as a default hyper value

- remove CWM (`sd_mecha.hypers.classes`)

- speedup `rotate` method by ~2x using `torch.svd_lowrank`

- add new builtin methods: `n_average`, `geometric_median`, `ties_sum_extended` / `add_difference_ties_extended`, `ties_sum_with_dropout` / `ties_with_dare`, `model_stock_for_tensor` / `model_stock_n_models`
- add a new parameter `vote_sgn` to `ties_sum` / `add_differenc_ties`

credits to 6DammK9 for this contribution!

- revert `sd_mecha.train_difference` to the original implementation from supermerger
- add 3 new methods `add_opposite`, `clamped_add_opposite` and `select_max_delta`

Note that `clamped_add_opposite` corresponds to the implementation of train difference in 0.0.19 and earlier versions.

- raise error for non-lora keys when loading a lora model type in https://github.com/ljleb/sd-mecha/pull/33
- fix a bug where keys of input models where directly used as fallback when applying a lora to a composite recipe in https://github.com/ljleb/sd-mecha/pull/34