SynthText dataset integration

[felixdittrich92]

This PR adds the Synthtext dataset

cf. #587

Problems:

• the unpack part needs a bit longer (>800k data) so i think we have to implement a check that unpacking is finished before continue
• same for the hash check this will end in a MemoryError

wdyt @fg-mindee how to handle? :)
suggestions are welcome 🤗

• tests
• bypass the hash check

[fg-mindee]

Thanks! I added some comments

[felixdittrich92]

@fg-mindee
Dataset size: ~41 GB

[fg-mindee]

Another round of comments 😅

[felixdittrich92]

@fg-mindee
I added to bypass the hash check and removed the tests for this dataset ftm also the train/val split now before iterating.
Already any decision about the dataset test ? :)

[felixdittrich92]

@fg-mindee
after attaching the example can you trigger integration again pls ? 😃

[fg-mindee]

Thanks for taking a look at this Felix!
I added some modification suggestions 👌

[fg-mindee]

Thanks, I added a few comments!

[fg-mindee]

Looks good, thanks!