feat: implement scanner similarity scoring

internal/scanner/scanner.go

@@ -0,0 +1,56 @@
+// Package scanner implements the core fuzzy-diff engine of NaviWatcher.
+//
+// It compares a user's local albums (from Navidrome) against an artist's
+// external discography (from MusicBrainz) and reports the releases that are
+// present externally but have no sufficiently similar local album.
+package scanner
+
+import (
+	"github.com/lithammer/fuzzysearch/fuzzy"
+	"naviwatcher/internal/normalize"
+)
+
+// Similarity returns a normalized similarity score in the range [0.0, 1.0]
+// between two strings. The strings are normalized first (lowercased,
+// bracketed/parenthesized content and years stripped, special characters
+// removed), then compared with a Levenshtein-distance-based ratio.
+//
+// A score of 1.0 means the normalized strings are identical; 0.0 means they
+// share nothing. Empty strings (after normalization) always score 0.0.
+func Similarity(a, b string) float64 {
+	na := normalize.NormalizeString(a)
+	nb := normalize.NormalizeString(b)
+
+	// Two empty inputs are not considered a match.
+	if na == "" && nb == "" {
+		return 0.0
+	}
+	// One empty, one non-empty: no similarity.
+	if na == "" || nb == "" {
+		return 0.0
+	}
+
+	dist := fuzzy.LevenshteinDistance(na, nb)
+	maxLen := len(na)
+	if len(nb) > maxLen {
+		maxLen = len(nb)
+	}
+
+	// Guard against maxLen == 0 (already handled above, but kept for safety).
+	if maxLen == 0 {
+		return 0.0
+	}
+
+	// 1.0 - normalized distance → higher is more similar.
+	score := 1.0 - float64(dist)/float64(maxLen)
+	if score < 0.0 {
+		return 0.0
+	}
+	return score
+}
+
+// IsMatch reports whether a and b are similar enough to be considered the
+// same release, given the provided threshold in [0.0, 1.0].
+func IsMatch(a, b string, threshold float64) bool {
+	return Similarity(a, b) >= threshold
+}