tasks: key derived captures by external id and record who resolved

A task extracted from mail deduped on the live-norm index only, so once he
finished it the row left the live set and the next poll of the same immutable
message re-extracted it as a fresh candidate. mavmaild is a read-only reader
and marks nothing read, so that repeats forever. Derived rows now carry an
ext_id built from the message uid and the extracted span, unique across every
status, while voice keeps live-only norm dedupe because saying an errand again
is the recurrence signal. A derived source can no longer capture straight to
open, and saying a task out loud that Maven had only proposed promotes the
candidate instead of answering that it is already in the list.

SetTaskStatus was classified AuthRead. Resolving a task is not additive, it
erases work off his list, so it is a write, and the row now records the caller
that moved it. ListTasks was unbounded. The list-query matcher claimed any
utterance with "что мне делать", including "с чем мне помочь", and the urgency
stripper matched inside words.

Found in review of #60.

internal/router/task.go

@@ -47,7 +47,10 @@ func ParseTaskCapture(text string) (TaskCapture, bool) {
 	rest := strings.TrimSpace(trimmed[len(best):])
 	rest = strings.TrimLeft(rest, ":—- ")
 	rest = strings.TrimSpace(rest)
-	rest = strings.TrimRight(rest, ".!")
+	// The question mark goes too. Whisper punctuates dictated Russian, and
+	// "добавь в задачи позвонить в банк?" must not store the mark or carry it
+	// into the dedupe key.
+	rest = strings.TrimRight(rest, ".!?")
 	rest, weight := stripUrgency(rest)
 	if rest == "" {
 		return TaskCapture{}, false
@@ -55,30 +58,86 @@ func ParseTaskCapture(text string) (TaskCapture, bool) {
 	return TaskCapture{Text: rest, Weight: weight}, true
 }
 
+// urgencyIntensifiers — words that may sit between the edge and the marker.
+// "очень срочно оплатить интернет" is the marker at the edge with one word in
+// front of it, and it means exactly what "срочно оплатить интернет" means.
+var urgencyIntensifiers = []string{"очень", "прям", "прямо", "really", "very", "super"}
+
+// urgencyEdgeTrim — punctuation to ignore around an edge token and to clean off
+// the remainder afterwards.
+const urgencyEdgeTrim = " .,;:!?—-"
+
 // stripUrgency pulls a leading or trailing urgency word out of the task text
 // and returns the weight it implies. Only at the edges: "срочно оплатить
 // интернет" and "оплатить интернет срочно" are the same instruction, while
 // "позвонить в срочную помощь" is a task whose text happens to contain the
 // stem, and cutting a word out of the middle of it would mangle the task.
 //
+// Matched as a TOKEN, not as a fixed prefix or suffix string. The old shape
+// required exactly one space before a trailing marker, so "оплатить интернет,
+// срочно" — which is what whisper produces from dictated Russian — kept weight
+// 0 and stored the comma and the word as part of the task, polluting the dedupe
+// key with the very flag he was trying to set.
+//
 // The word is removed from the text, because the list should read "оплатить
 // интернет (важно)" and not "важно оплатить интернет (важно)".
 func stripUrgency(text string) (string, int) {
+	fields := strings.Fields(text)
+	if len(fields) == 0 {
+		return text, 0
+	}
 	for _, m := range urgencyMarkers {
-		lower := strings.ToLower(text)
-		switch {
-		case strings.HasPrefix(lower, m.Word+" "):
-			return strings.TrimSpace(text[len(m.Word):]), m.Weight
-		case strings.HasSuffix(lower, " "+m.Word):
-			return strings.TrimSpace(text[:len(text)-len(m.Word)]), m.Weight
-		case lower == m.Word:
-			// Nothing but the marker — no task in it.
-			return "", 0
+		// Strongest marker first (task_phrases.go sorts them), leading edge
+		// before trailing, so a text carrying both keeps the stronger one.
+		if lo, hi, ok := urgencySpan(fields, m.Word); ok {
+			rest := strings.Join(append(append([]string{}, fields[:lo]...), fields[hi+1:]...), " ")
+			rest = strings.Trim(rest, urgencyEdgeTrim)
+			if rest == "" {
+				// Nothing but the marker — no task in it.
+				return "", 0
+			}
+			return rest, m.Weight
 		}
 	}
 	return text, 0
 }
 
+// urgencySpan finds the marker at either edge, allowing intensifiers between
+// the edge and the marker, and returns the inclusive token range to cut.
+func urgencySpan(fields []string, word string) (lo, hi int, ok bool) {
+	for i := 0; i < len(fields); i++ {
+		if isUrgencyToken(fields[i], word) {
+			return 0, i, true
+		}
+		if !isIntensifier(fields[i]) {
+			break
+		}
+	}
+	for i := len(fields) - 1; i >= 0; i-- {
+		if isUrgencyToken(fields[i], word) {
+			return i, len(fields) - 1, true
+		}
+		if !isIntensifier(fields[i]) {
+			break
+		}
+	}
+	return 0, 0, false
+}
+
+func isUrgencyToken(tok, word string) bool {
+	return strings.Trim(strings.ToLower(tok), urgencyEdgeTrim) == word
+}
+
+func isIntensifier(tok string) bool {
+	t := strings.Trim(strings.ToLower(tok), urgencyEdgeTrim)
+	for _, w := range urgencyIntensifiers {
+		if t == w {
+			return true
+		}
+	}
+	return false
+}
+
 // IsTaskListQuery reports whether an utterance asks for the outstanding task
 // list — "какие у меня задачи?", "что мне нужно сделать?", "список дел".
 //
@@ -94,13 +153,23 @@ func IsTaskListQuery(text string) bool {
 	if hasTok(toks, "как") && (hasTok(toks, "дела") || hasTok(toks, "делишки")) {
 		return false
 	}
-	// "что мне нужно сделать" / "что мне делать" — no task noun at all.
-	if (hasTok(toks, "что") || hasTok(toks, "чем")) &&
-		(hasTok(toks, "сделать") || hasTok(toks, "заняться")) {
-		return true
-	}
-	if hasTok(toks, "what") && hasTok(toks, "do") {
-		return true
+	// "что мне нужно сделать" / "чем мне заняться" — no task noun at all, so
+	// the pronoun is what carries the meaning. Without it these rules claimed
+	// every question with a verb in them: "что нужно сделать чтобы перезапустить
+	// сервер?" and "what does docker do?" both answered "задач нет." from ahead
+	// of recall and the model, which is the failure the source ordering exists
+	// to avoid, pointed the other way.
+	//
+	// A "с"/"со" object excludes them too: "что мне сделать с этим файлом" has
+	// the pronoun and is still a question about a file.
+	if !hasTok(toks, "с") && !hasTok(toks, "со") {
+		if hasTok(toks, "мне") && (hasTok(toks, "что") || hasTok(toks, "чем")) &&
+			(hasTok(toks, "сделать") || hasTok(toks, "делать") || hasTok(toks, "заняться")) {
+			return true
+		}
+		if hasTok(toks, "what") && hasTok(toks, "do") && hasTok(toks, "i") && !hasTok(toks, "you") {
+			return true
+		}
 	}
 	hasNoun := false
 	for _, t := range toks {