Faster image pipelines

Some low-hanging fruit that FastAI.jl can pick to become faster here:

- builtin presizing support
- prototype at https://github.com/lorenzoh/DataBlockStorage.jl/
- using JpegTurbo.jl's downsized loading so no storing to disk is needed (https://github.com/JuliaIO/JpegTurbo.jl/issues/15#issuecomment-1031548344)
- use inplace operations
- DataAugmentation.jl already supports inplace operations, FastAI.jl just needs to add a buffered `encode!` interface and use those paths
- JpegTurbo.jl image loading could also be done inplace (https://github.com/JuliaIO/JpegTurbo.jl/issues/23)
- the whole pipeline could become non-allocating
- use MLUtils.jl (#196)
- new `DataLoader` seems more stable and possibly a little faster
- run projective transformations on GPU (https://github.com/lorenzoh/DataAugmentation.jl/issues/48)