Add support for MasakhaPOS Dataset

[stefan-it]

Hi,

this PR adds support for the recently proposed MasakhaPOS Dataset.

Details can be found in this tweet.

The dataset is available in this repo: https://github.com/masakhane-io/masakhane-pos

I received preprint of the paper and wrote unit tests to check number of parsed sentences for dataset splits for each language.

Example usage of MasakhaPOS:

from flair.datasets import MASAKHA_POS

corpus = MASAKHA_POS(languages="bam")

[stefan-it]

/cc @dadelani

[stefan-it]

@alanakbik Please let me know, if dataset name is ok: it does not quite match into the UD_ naming scheme, because dataset is not in Universal Dependencies format, as it only has token and upos.

[dadelani]

@stefan-it , the dataset name is MasakhaPOS, arXiv paper will be out tomorrow

[stefan-it]

Thanks @dadelani for feedback, I corrected the dataset name now :)

[stefan-it]

Preprint is now available here 🤗

[stefan-it]

Hi @helpmefindaname do you accidentally have an idea, why poetry stage will get stuck in timeout:

https://github.com/flairNLP/flair/actions/runs/5062453919/jobs/9099744358?pr=3247

This was the already the case yesterday, I've just re-ran the build, but same error.

[helpmefindaname]

Hi @helpmefindaname do you accidentally have an idea, why poetry stage will get stuck in timeout:

https://github.com/flairNLP/flair/actions/runs/5062453919/jobs/9099744358?pr=3247

This was the already the case yesterday, I've just re-ran the build, but same error.

The dependency resolution took needlessly long, as it tried out all 300+ boto3 versions with an incompatible dependency before taking the right one.
#3249 should fix the problems.

[stefan-it]

After #3258 I will do a rebase now :)

[alanakbik]

@stefan-it thanks for adding this! And thanks @dadelani for creating this dataset!