Use torchvision transforms instead of PIL operations for solarization and gaussian blurring

Nobody has claimed this yet.

**Problem**
When using a transform that uses `lightly.transforms.gaussian_blur.GaussianBlur` or `lightly.transforms.solarization.RandomSolarization`, e.g. `DINOTransform`, the input must be transformed to a `PIL.Image` first, because the Gaussian blur and the solarization use `PIL.ImageFilter.GaussianBlur` and `PIL.ImageOps.solarize` respectively. For training, the `PIL.Image` has to be transformed back to a tensor.

**Background**
It might be beneficial to get rid of this conversion for performance purposes. Sometimes, using a different reader like pyvips/openslide/cucim is necessary to extract only a patch of an image because images are too big to fit in memory (e.g. in computational pathology or remote sensing). Here, often patches are extracted and immediately transformed to torch tensors for torchvision to transform the patches. Doing all the transforms on tensors overcomes converting to other formats, like PIL, increasing performance.

**Alternative**
Torchvision provides seemingly fast (jitted) implementations for solarization [1] and gaussian blurring [2].

I'm not aware of any other transforms in the lightly package that rely on PIL.

**References**
[1] [implementation](https://github.com/pytorch/vision/blob/4249b610811b290ea9ac9e445260be195ce52ae1/torchvision/transforms/v2/functional/_color.py#L483-L486) [docs](https://pytorch.org/vision/main/generated/torchvision.transforms.v2.RandomSolarize.html#torchvision.transforms.v2.RandomSolarize)
[2] [implementation](https://github.com/pytorch/vision/blob/4249b610811b290ea9ac9e445260be195ce52ae1/torchvision/transforms/v2/functional/_misc.py#L75-L78) [docs](https://pytorch.org/vision/main/generated/torchvision.transforms.v2.GaussianBlur.html?highlight=gaussianblur#torchvision.transforms.v2.GaussianBlur)

Start with lightly.transforms.gaussian_blur.GaussianBlur and lightly.transforms.solarization.RandomSolarization, then trace how DINOTransform uses them. Compare their PIL-based operations with the referenced torchvision implementations. Done means solarization and Gaussian blurring work on tensors without the described PIL conversions while preserving the existing transform behavior.

Written by the indexing model from the issue text.