Full correctness check against the docs: safer writes, reverts and texts

Safety
- Writes are serialized (Workspace.lock): two background jobs rebuilding one
  package (a sound and its subtitle applied side by side) lost an edit.
- A store copies a bank file or package as "original" only if it matches the
  shipped catalog (Workspace.unoriginal_files: in-memory sounds, bank copies of
  voice lines, subtitles); the editor refuses to switch stores while changes
  are applied. A second store could otherwise resurrect a reverted sound.
- Mod import in the editor asks before overwriting files changed outside the
  tool (it forced silently).
- Revert all in the editor reverts texts too; the uninstaller counts text
  edits before offering to delete the store.
- ackit restore-backup: the full-backup restore decisions.md described; it
  keeps the sounds of WADs without a full copy recorded.
- build_catalog also stops when a subtitle differs from the catalog.

Languages and texts
- Game.language() returns languages without voices (JPN, RUS, ...) and
  Game.voice_archive() their INT voices; ackit text --lang JPN works.
- Menu lookups parse one cached file (find & replace: minutes -> ms).
- Subtitles decode cp1252 as Windows does: the UTF-8-in-cp1252 line
  Speech_TAPE_J_Answer_1_00 (KOR/POL/POR/RUS) is written back byte-for-byte;
  SUB_KOR/SUB_RUS rebuilt from the original game (only those rows changed).
- Mods exported while another language is loaded keep name and line.

Other fixes
- Down-mix folds extra channels in at -3 dB (a 5.1 centre voice was dropped).
- e2e_editor no longer overwrites the user's settings.json; kill_game waits
  for the game to exit before reverts.
- Recorder handles Int32/UInt8 input and split frames; package block cache
  locked; conflict messages, CLI help examples, stale docstrings and doc
  references fixed; tests no longer read this PC's game language.

Docs (decisions, user guide, README, subtitles.md, banks.md, catalog README,
REPORT, NOTES, CLAUDE.md) updated; check_docs: 102 claims match. 148 tests;
e2e_features 9/9, e2e_editor and verify PASS; game original afterwards.

app/audioio.py

@@ -163,7 +163,8 @@ class Recorder(QObject):
         self.chunks = []
         self.rate = 48000
         self.channels = 1
-        self.fmt_float = False
+        self.sample = "<i2"            # numpy type of the device's samples
+        self._rest = b""               # bytes of an incomplete frame, completed by the next read
         self._sim = None
 
     def set_device(self, device):
@@ -173,7 +174,7 @@ class Recorder(QObject):
         return self.source is not None or self._sim is not None
 
     def start(self):
-        self.chunks = []
+        self.chunks, self._rest = [], b""
         dev = self.device if (self.device is not None and not self.device.isNull()) else (
             QMediaDevices.defaultAudioInput() if audio_enabled() else None)
         if dev is None or dev.isNull():
@@ -186,7 +187,8 @@ class Recorder(QObject):
         if not dev.isFormatSupported(fmt):
             fmt = dev.preferredFormat()
         self.rate, self.channels = fmt.sampleRate(), fmt.channelCount()
-        self.fmt_float = fmt.sampleFormat() == QAudioFormat.SampleFormat.Float
+        F = QAudioFormat.SampleFormat
+        self.sample = {F.Float: "<f4", F.Int32: "<i4", F.UInt8: "u1"}.get(fmt.sampleFormat(), "<i2")
         self.source = QAudioSource(dev, fmt, self)
         self.io = self.source.start()
         if self.io is None:
@@ -202,13 +204,20 @@ class Recorder(QObject):
         self._ingest(bytes(self.io.readAll()))
 
     def _ingest(self, raw):
+        raw = self._rest + raw
+        frame = np.dtype(self.sample).itemsize * self.channels
+        n = len(raw) // frame * frame
+        raw, self._rest = raw[:n], raw[n:]
         if not raw:
             return
-        if self.fmt_float:
-            x = np.frombuffer(raw, dtype="<f4")
-        else:
-            x = np.frombuffer(raw[: len(raw) // 2 * 2], dtype="<i2").astype(np.float32) / 32768.0
-        x = x[: len(x) // self.channels * self.channels].reshape(-1, self.channels)
+        x = np.frombuffer(raw, dtype=self.sample)
+        if self.sample == "<i2":
+            x = x.astype(np.float32) / 32768.0
+        elif self.sample == "<i4":
+            x = (x / 2147483648.0).astype(np.float32)
+        elif self.sample == "u1":
+            x = (x.astype(np.float32) - 128.0) / 128.0
+        x = x.reshape(-1, self.channels)
         self.chunks.append(x)
         self.level.emit(float(np.max(np.abs(x))) if len(x) else 0.0)
         self.elapsed.emit(sum(len(c) for c in self.chunks) / float(self.rate))
@@ -216,8 +225,8 @@ class Recorder(QObject):
     # simulated input (tests / no microphone) ----------------------------------------------------
     def start_simulated(self, x, rate, realtime=True):
         """Feed `x` as if it came from a microphone (20 ms blocks)."""
-        self.chunks = []
-        self.rate, self.channels, self.fmt_float = rate, x.shape[1], True
+        self.chunks, self._rest = [], b""
+        self.rate, self.channels, self.sample = rate, x.shape[1], "<f4"
         blocks = [x[i:i + rate // 50] for i in range(0, len(x), rate // 50)]
         if not realtime:
             for b in blocks: