fix: the Yaesu keyer refused to send — a status reply we read too strictly

"the keyer buffer stayed full for 4s" and nothing keyed, on a rig whose CAT is
working. The buffer was not full: the check demanded the reply be exactly "KY0;"
with the digit at byte 2, so anything the FTDX10 phrases differently — a space
before the digit, another command's reply arriving first — read as "still full"
on every poll until the deadline.

The test is now asymmetric on purpose: only a clear "1" holds the send back.
Anything unrecognised goes ahead. Refusing to transmit because a status line was
phrased unexpectedly is the worse failure — the operator gets silence with no
explanation, where at worst sending early truncates a long message, which the
chunk loop then recovers from.

The reply is also logged once per run, since we have no verified sample of it —
that line is what turns the next surprise into a fact instead of a theory.

internal/cat/yaesu_cw.go

@@ -27,6 +27,10 @@ import (
 // split and fed as the rig drains its buffer.
 const yaesuCWChunk = 24
 
+// cwStatusLogged makes the KY status reply appear in the log once per run: it
+// is a line we have no verified sample of, so the first one is worth keeping.
+var cwStatusLogged bool
+
 // yaesuCWAllowed is what the rig's keyer can actually send. Anything else is
 // dropped rather than passed through: an unsupported byte can abort the whole
 // buffer, losing the rest of the message with no error anywhere.
@@ -97,16 +101,57 @@ func (y *Yaesu) waitCWBuffer(within time.Duration) error {
 			debugLog.Printf("yaesu cw: KY status query failed (%v) — sending anyway", err)
 			return nil
 		}
-		if len(r) >= 3 && r[2] == '0' {
-			return nil // room in the buffer
+		// Only an explicit "full" holds us back.
+		//
+		// The first version required the reply to be exactly "KY0;" at byte 2, and
+		// on a real FTDX10 that refused to send at all: anything the rig phrases
+		// differently — a space before the digit, a longer status — read as "still
+		// full" and the send died after four seconds having keyed nothing. The rig
+		// is the one that knows; when its answer is not a clear "1", the right move
+		// is to go ahead rather than to sit and time out.
+		// Log the shape once per session: this is a reply we have no verified
+		// sample of, and the log is what turns the next surprise into a fact.
+		if !cwStatusLogged {
+			cwStatusLogged = true
+			debugLog.Printf("yaesu cw: KY status reply is %q (full=%v)", r, yaesuCWBufferFull(r))
+		}
+		if !yaesuCWBufferFull(r) {
+			return nil
 		}
 		if time.Now().After(deadline) {
-			return fmt.Errorf("yaesu: the keyer buffer stayed full for %s", within)
+			return fmt.Errorf("yaesu: the keyer buffer stayed full for %s (last reply %q)", within, r)
 		}
 		time.Sleep(50 * time.Millisecond)
 	}
 }
 
+// yaesuCWBufferFull reads the KY status reply.
+//
+// Deliberately asymmetric: only a clear "1" means full. Anything else — an
+// unexpected shape, a reply from another command that arrived first, an empty
+// string — is treated as "go ahead". Refusing to send because a status line was
+// phrased unexpectedly is the worse failure: the operator gets silence with no
+// explanation, where at worst sending early truncates a long message.
+func yaesuCWBufferFull(reply string) bool {
+	r := strings.TrimSpace(reply)
+	if !strings.HasPrefix(strings.ToUpper(r), "KY") {
+		return false
+	}
+	for _, c := range r[2:] {
+		switch c {
+		case '0':
+			return false
+		case '1':
+			return true
+		case ' ', ';':
+			continue
+		default:
+			return false // not a status digit — don't block on it
+		}
+	}
+	return false
+}
+
 // filterYaesuCW upper-cases and strips what the keyer cannot send.
 func filterYaesuCW(text string) string {
 	var b strings.Builder