Translation on Firefox Translations models; new logo

- Email translation moves from LibreTranslate to our own translator
  service (translator/): Mozilla's Firefox Translations models on the
  Bergamot engine, fastText language detection (176 languages). Polish,
  English, German, French, Spanish and Italian on first start (~350 MB
  instead of ~10 GB); Admin panel > Email translation installs and
  removes languages. Built and pushed as the `translator` image.
- Every message is translated paragraph by paragraph as plain text;
  names travel with their sentence and are kept only where the model
  changed them; a sentence start the model copied ("Enclosed") is tried
  again in lower case; list markers are not sent
- The Translate button shows only for installed languages
- New logo: favicon, app/maskable/iPhone icons, notification badge,
  email logo and calendar copy regenerated; BIMI logo rebuilt in SVG
  Tiny PS with its gradients; masks render in browsers
- Launch screen background is the app's dark grey

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

backend/app/translation_prep.py

@@ -1,25 +1,25 @@
 """Getting a message ready for translation, and tidying what comes back.
 
-LibreTranslate translates each piece of text between tags on its own, and its
-small models stumble on things a reader would not: a sentence wrapped onto two
-lines becomes two sentences ("ask you about" / "Plan for next month"),
-"Best," becomes "Najlepsze," and a company name gets translated word by word.
-So before a message goes out:
+A message is translated as plain text, piece by piece (the text between
+tags), and the models stumble on things a reader would not: a sentence
+wrapped onto two lines becomes two sentences ("ask you about" / "Plan for next
+month"), "Best," becomes "Najlepsze," and a company name gets translated word
+by word. So before a message goes out:
 
 - lines that carry on a sentence (no end punctuation, next starts lowercase)
   are joined, in HTML (<br>, one <div> per line) and in wrapped plain text;
 - greetings and sign-offs that make a whole line are put in the target
   language from a small phrase table (English, Polish, German);
-- names are marked translate="no", which LibreTranslate keeps — the people
-  on the message, the user's own names and signature names, and capitalised
-  multi-word names mid-sentence — but only the names the model would change
-  ("Milan" → "Mediolan"): a kept name splits its sentence, and LibreTranslate
-  translates each piece alone ("Dear <kept>Lukas Remis</kept>," →
-  "Kochanie."), so a name the model leaves alone stays in the sentence;
+- names are marked translate="no" and not sent — the people on the message,
+  the user's own names and signature names, and capitalised multi-word names
+  mid-sentence — but only the names the model would change ("Milan" →
+  "Mediolan"): a kept name splits its sentence into pieces translated alone
+  ("Dear <kept>Lukas Remis</kept>," → "Kochanie."), so a name the model
+  leaves alone stays in the sentence;
 - so are emoji, file names ("A_0023C017_p1CAC.mov") and text without a letter;
-- a message in more than one language is translated paragraph by paragraph,
-  each from its own language (split_pieces); those in the target stay;
-- and afterwards, the space LibreTranslate drops after a kept word comes back.
+- the message is translated paragraph by paragraph (split_pieces), each from
+  its own language when there are several; those in the target stay;
+- and afterwards, a kept word gets back the spaces around it.
 """
 
 from __future__ import annotations
@@ -315,7 +315,7 @@ def split_pieces(html: str) -> tuple[list[str], list[Paragraph]]:
 
 # --- after translation -------------------------------------------------------------
 
-# LibreTranslate drops the space after a kept span ("Bruton</span>i twoje").
+# A kept span can lose the space beside it ("Bruton</span>i twoje").
 # Only kept spans: they still carry translate="no" here (the sanitizer that
 # runs next takes it off); other spans are the sender's and stay as they are.
 _KEPT = re.compile(r'(<span translate="no">[^<]*</span>)')