fix(flow): drain trailing audio and harden chunked ASR pipeline

- P0/P1: tail-drain state machine in FlowContinuousCapture with converter
  flush; FlowSessionManager awaits drain before finalize
- P2: FlowCaptureTailDrain policy/tracker, pipeline diagnostics, unit tests
- P3: short final chunk merged re-transcription in ChunkedUtterancePipeline
- P4: UtteranceTranscriptStitcher composedSafely fallback; preview path parity
  in LiveDictationController

Co-authored-by: Rocky <hkgood@users.noreply.github.com>

OSGKeyboardTests/UtteranceTranscriptStitcherTests.swift

@@ -20,4 +20,24 @@ final class UtteranceTranscriptStitcherTests: XCTestCase {
         stitcher.append(index: 0, text: "第一段")
         XCTAssertEqual(stitcher.composed(), "第一段 第二段")
     }
+
+    func testComposedSafelyFallsBackWhenOverlapMergeShortensTooMuch() {
+        var stitcher = UtteranceTranscriptStitcher()
+        stitcher.append(index: 0, text: "今天天气很好我们")
+        stitcher.append(index: 1, text: "去公园")
+        let merged = stitcher.composed()
+        let safe = stitcher.composedSafely()
+        XCTAssertFalse(merged.isEmpty)
+        XCTAssertFalse(safe.isEmpty)
+        XCTAssertTrue(safe.contains("去公园"))
+    }
+
+    func testRemoveLastSegmentSupportsMergedTailRetranscription() {
+        var stitcher = UtteranceTranscriptStitcher()
+        stitcher.append(index: 0, text: "第一段")
+        stitcher.append(index: 1, text: "第二段")
+        stitcher.removeLastSegment()
+        stitcher.append(index: 1, text: "第二段合并")
+        XCTAssertEqual(stitcher.composed(), "第一段 第二段合并")
+    }
 }