[pt] Migrate spelling dictionaries to Morfologik format

[jaumeortola]

Migrate PT dictionary to Morfologik

This PR:

• removes the old Hunspell data;
• uses Morfologik-format binary dictionary data hosted on Sonatype;
• adds a bunch of tests (mostly speller and tokeniser);
• moves a lot (not all) of the data from hunspell/spelling.txt to spelling.txt;
• reworks the tokeniser to use the same break characters as the other locales.

The work with curation of the dictionary source data was done in the portuguese-pos-dict repo.

To do before merging

• compare this branch vs. master with a large corpus;
• resolve problems (caused by tokenization, and others);
• release the new dictionaries;
• remove old Hunspell files and rules;
• add more tests;
• add words to resource/pt/do_not_suggest.txt.

---

@jaumeortola , I've taken all the steps to release it. As far as I can see, it's live. But the tests still fail when trying to read the contents of the live repository. This part of it I have no idea about.

@danielnaber , you can already review the Java, I don't reckon it'll change.

@susanaboatto ... lol yeah there's a lot of stuff. Do have a look at the changes made to the disambiguator and the new files (spelling.txt, do_not_suggest.txt, and dialect_alternations.txt). All the other PT-specific changes are really made directly to the dictionary source data in the PT dictionary repo.

---

^{Before merging, please let me squash this into something acceptable.}

[danielnaber]

But the tests still fail when trying to read the contents of the live repository. This part of it I have no idea about.

I restarted the CI process and this time it has worked, it seems.

[p-goulart]

@SteVio89 this is the PR that @maphjo mentioned yesterday with the partial test deploy for PT dictionaries.

[p-goulart]

@SteVio89 the rule ID we had was not being underlined in red – I thought it was enough for it to contain MORFOLOGIK but it had to contain MORFOLOGIK_RULE. So I've pushed new code with the updated rule ID and, once the tests pass, can we deploy that on the test server? Sorry about that.

[p-goulart]

Cannot merge until we work out how to get usage data from partial rollout.

[p-goulart]

After yesterday's test deployment:

• I've checked the results on Matomo to make sure we're getting proper data;
• the apply rate of the new rule hovers around 60%; the Hunspell version had ~55%:
  

I think it's safe to say we can merge this! Big success 🥳

Thanks @jaumeortola @maphjo @SteVio89 for the support with this endeavour!

[danielnaber]

5 more percentage points is great! Thanks to everyone who has been working on this!