Fix: Add migration to use correct encoding from environment

* The issue was fixed in code in revisions 4c89be6b and a82dc686. This
  commit uses the same approach, but fixes the existing malformed records
  in the users table in an alembic migration during upgrade.
* First, an approach based on plpython3u extension and a user-defined
  function was considered, which had the downside of added dependency.
  This one is implemented in Python - data is read from the database,
  the demangling is performed and the correct new data is UPDATEd into
  the database.

(Redmine issue: #7553)

lib/hawat/migrations/versions/1c380ebe027e_demangle_national_characters_in_auth_.py

+"""Demangle national characters in auth attributes
+
+Revision ID: 1c380ebe027e
+Revises: 432c8bc8e49b
+Create Date: 2022-07-25 12:38:15.843315
+
+"""
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision = '1c380ebe027e'
+down_revision = '432c8bc8e49b'
+branch_labels = None
+depends_on = None
+
+
+def demangle(value):
+    """National character demangling helper"""
+    try:
+        return value.encode("iso-8859-1").decode()
+    except (UnicodeDecodeError, UnicodeEncodeError):
+        return value
+
+
+def upgrade():
+    conn = op.get_bind()
+    rows = conn.execute('SELECT id, fullname, organization FROM users').fetchall()
+    for rid, fnm_old, org_old in rows:
+        fnm = demangle(fnm_old)
+        org = demangle(org_old)
+        if fnm != fnm_old or org != org_old:
+            conn.execute("UPDATE users SET fullname = %s, organization = %s WHERE id = %s", (fnm, org, rid))
+
+
+def downgrade():
+    pass