reorder training preprocessing modules in extras tab

[w-e-w]

fix issue #14360 Captions not being generated for flipped copies
since captioning is applied before flipping the image the flip the version of the image don't have captions

I reordered the five image training preprocessing modules using the order from before the rework 11d23e8 remove Train/Preprocessing tab and put all its functionality into extras batch images mode
(hopefully I didn't mess up the order)

current order	new default

---

personally I do not like 11d23e8 remove Train/Preprocessing tab and put all its functionality into extras batch images mode
I think the use case of preprocessing images for training and applying additional effects such as upscaling among other things that extensions do such as color enhancement are very different and should be separated
lots of people are not interested in training (or use other tools for training) would probably never uses preprocessing, for them this will be just cluttering the UI
the modules also cannot be disabled / hidden easily

personally I like the version before the rework, but I will have less of an issue if all these is grouped under a "big accordion"

---

there's also another issue caused by the rework

• captioning is now extremely slow compared to what it was before
  irrc before captioning was done in a single large batch as opposed to being processed separately
  which means that you can just load and unload the model once reducing overhead

assuming that this is the only cause of the slowdown, then adding a function before_batch_processing and after_batch_processing can solve this issue, they can be used to load it unload the model

Checklist:

• I have read contributing wiki page
• I have performed a self-review of my own code
• My code follows the style guidelines
• My code passes tests