Preserve \n through whitespace cleanup so "new line" survives strip

The "new line" → \n substitution happens before the final cleanup. When
a VAD segment contains only "new line", the resulting lone \n was wiped
by text.strip() and the surrounding \s+ patterns, so the segment came
back as "" and the client discarded it. Switched the whitespace patterns
to [ \t] / [ \t\r] so \n survives to the client.

server.py

@@ -78,10 +78,15 @@ def _replace_spoken_punctuation(text: str) -> str:
     text = re.sub(r"\{([,.:;!?/\-+])\}", r"\1", text)
     for pattern, replacement in _SPOKEN_PUNCTUATION:
         text = pattern.sub(replacement, text)
-    text = re.sub(r"\s+([,.:;!?)\]])", r"\1", text)
-    text = re.sub(r"([([\[])\s+", r"\1", text)
+    # NB: \s would also match \n that we just inserted via "new line" — use
+    # [ \t] so the newline survives to the client.
+    text = re.sub(r"[ \t]+([,.:;!?)\]])", r"\1", text)
+    text = re.sub(r"([([\[])[ \t]+", r"\1", text)
     text = re.sub(r"  +", " ", text)
-    return text.strip()
+    # Trim spaces/tabs/CR only — preserving \n means a segment that's just
+    # "new line" (now "\n") doesn't get wiped to empty and discarded by the
+    # client as a zero-length segment.
+    return re.sub(r"^[ \t\r]+|[ \t\r]+$", "", text)
 
 
 # ---------------------------------------------------------------------------