feat: real conversation threading at ingest + migration script

The IMAP parser only extracted Message-ID, so In-Reply-To / References
were never persisted and threadId stayed empty -> conversations never
grouped with real data (verified: 162/162 empty in user's DB).

- ParsedMimeMessage: add inReplyTo + references fields
- MimeStorage::parseMessage: extract In-Reply-To and References headers
- ImapSynchronizer parseFetchResponseBytes: propagate inReplyTo/references,
  compute + persist threadId at ingest time (ThreadUtils::extractThreadId)
- MimeStorage::storeMessage: propagate the same headers on MailItem
- scripts/rethread.py: migrate existing DBs by re-reading .eml files
  (recovered In-Reply-To/References from raw MIME and recomputed threadId:
  149/162 grouped, incl. Nacho x6, Bitar x4, Mariel x3, etc.)

Verified against user's real DB copy.

scripts/rethread.py

@@ -0,0 +1,96 @@
+#!/usr/bin/env python3
+"""Recupera In-Reply-To/References del .eml y recalcula threadId en Wino Mail.
+Uso: python3 rethread2.py <bd.sqlite> <carpeta_mails>
+"""
+import sys, re, sqlite3, os, email
+
+bd = sys.argv[1]
+mails_dir = sys.argv[2]
+conn = sqlite3.connect(bd)
+cur = conn.cursor()
+
+cur.execute("SELECT id, messageId FROM MailCopy")
+rows = cur.fetchall()
+
+# indexar emls por message-id en nombre de archivo
+eml_by_mid = {}
+for root, _, files in os.walk(mails_dir):
+    for f in files:
+        if not f.endswith(".eml"):
+            continue
+        # nombre base contiene el message-id antes de .eml
+        base = f[:-4]
+        eml_by_mid[base] = os.path.join(root, f)
+
+def parse_headers(p):
+    try:
+        msg = email.message_from_bytes(open(p, "rb").read())
+    except Exception:
+        return None
+    mid = msg.get("Message-ID", "")
+    irt = msg.get("In-Reply-To", "")
+    refs = msg.get("References", "")
+    return mid, irt, refs
+
+def thread_id(mid, irt, refs):
+    if refs:
+        m = re.search(r"<([^>]+)>", refs)
+        return m.group(1) if m else refs.strip("<>")
+    if irt:
+        m = re.search(r"<([^>]+)>", irt)
+        return m.group(1) if m else irt.strip("<>")
+    if mid:
+        m = re.search(r"<([^>]+)>", mid)
+        return m.group(1) if m else mid.strip("<>")
+    return ""
+
+upd_thread = 0
+upd_irt = 0
+upd_refs = 0
+not_found = 0
+no_headers = 0
+
+for rid, mid in rows:
+    mid = (mid or "").strip()
+    # buscar eml: por nombre base == mid, o mid contenido
+    p = None
+    if mid in eml_by_mid:
+        p = eml_by_mid[mid]
+    if p is None:
+        # intentar coincidencia parcial (dominio distinto)
+        mid_nodot = mid.split("@")[0][:20]
+        for base, path in eml_by_mid.items():
+            if mid_nodot and mid_nodot in base:
+                p = path
+                break
+    if p is None:
+        not_found += 1
+        continue
+
+    hdrs = parse_headers(p)
+    if hdrs is None:
+        not_found += 1
+        continue
+    emid, eirt, erefs = hdrs
+
+    tid = thread_id(emid or mid, eirt, erefs)
+    # guardar headers recuperados
+    cur.execute("UPDATE MailCopy SET inReplyTo=?, \"references\"=?, threadId=? WHERE id=?",
+                (eirt.strip() if eirt else "", erefs.strip() if erefs else "", tid or "", rid))
+    if tid: upd_thread += 1
+    if eirt.strip(): upd_irt += 1
+    if erefs.strip(): upd_refs += 1
+
+conn.commit()
+print(f"correos: {len(rows)}")
+print(f"eml no encontrados: {not_found}")
+print(f"inReplyTo recuperado: {upd_irt}")
+print(f"references recuperado: {upd_refs}")
+print(f"threadId asignado: {upd_thread}")
+
+# agrupar en hilos
+print("\n=== hilos detectados (threadId con >1 correo) ===")
+cur.execute("SELECT threadId, COUNT(*), MAX(sender) FROM MailCopy WHERE threadId!='' GROUP BY threadId HAVING COUNT(*)>1 ORDER BY COUNT(*) DESC LIMIT 15")
+for row in cur.fetchall():
+    print(row)
+conn.close()