Add support for more Latin UD datasets

[stefan-it]

Hi,

this PR adds some more Latin UD datasets:

• UD_Latin-CIRCSE
• UD_Latin-ITTB
• UD_Latin-Perseus
• UD_Latin-PROIEL
• UD_Latin-UDante

This feature request came up in #3391.

Example

The datasets can be loaded, e.g. with:

from flair.datasets import UD_LATIN_CIRCSE, UD_LATIN_ITTB, UD_LATIN_UDANTE, UD_LATIN_PERSEUS, UD_LATIN_PROIEL

# UD_LATIN_CIRCSE
corpus_circse = UD_LATIN_CIRCSE()
str(corpus_circse)
# Outputs: Corpus: 0 train + 0 dev + 1263 test sentences

# UD_LATIN_ITTB
corpus_ittb = UD_LATIN_ITTB()
str(corpus_ittb)
# Outputs: Corpus: 22775 train + 2101 dev + 2101 test sentences

# UD_LATIN_UDANTE
corpus_udante = UD_LATIN_UDANTE()
str(corpus_udante)
# Outputs: Corpus: 926 train + 376 dev + 421 test sentences

# UD_LATIN_PERSEUS
corpus_perseus = UD_LATIN_PERSEUS()
str(corpus_perseus)
# Outputs: Corpus: 1334 train + 0 dev + 939 test sentences

# UD_LATIN_PROIEL
corpus_proiel = UD_LATIN_PROIEL()
str(corpus_proiel)
# Outputs: Corpus: 16196 train + 1233 dev + 1260 test sentences

Unittests

For all newly added datasets and the UD_LATIN dataset, unit tests were written, to check if the number of sentences match the reported number of sentences from the UD stats.

Overall Stats

Here are some overall stats of all supported UD datasets for Latin in Flair so far:

Dataset	# Train sentences	# Dev Sentences	# Test Sentences	# Total Sentences
UD_LATIN	7,289	850	884	9,023
UD_LATIN_CIRCSE	0	0	1,263	1,263
UD_LATIN_ITTB	22,775	2,101	2,101	26,977
UD_LATIN_UDANTE	926	376	421	1,723
UD_LATIN_PERSEUS	1,334	0	939	2,273
UD_LATIN_PROIEL	16,196	1,233	1,260	18,689