DeepL: added Glossary and HTML functionality

[thornch]

DeepL: Added support for the parameters tag_handling and glossary_id via application.config:

"translations.deepl.tagHandling": "html" // Enables translation of HTML content, e.g. from rich text editors.
"translations.deepl.glossaryById": "{client-glossary-id}" // Uses a specific DeepL glossary ID for translations.
"translations.deepl.glossaryByName": "{client-glossary-name}" // Resolves the glossary ID by glossary name before translation.

DeepL glossary handling:
When a glossary is replaced or recreated, its glossary ID may change. To keep glossary functionality stable in such cases, keep the glossary name unchanged and configure glossaryByName instead of glossaryById.

If a glossary name is configured, the glossary lookup is performed using the glossary name together with the source and target language pair.

The resolved glossary ID is cached in memory. If the glossary becomes invalid (for example because it was deleted or recreated), the cache is invalidated when a translation request fails and the glossary ID is resolved again on the next request.

Note: DeepL requires source_lang to be specified explicitly when a glossary is used for translation.

[thornch]

All code review comments and requested changes from #93 have been resolved and implemented as part of this pull request.

[SebastianStehle]

What exactly is this cache and how much entries can we expect here? I have not digged into this glossary stuff.