feat(web): /tags tabs + keyword surfacing + dup tab restyle

- /tags hosts four tabs (Labels (auto) / Keywords / Ratings / Colors),
  URL-driven with pagination on the label + keyword grids; ratings and
  colors stay as fixed buckets.
- /duplicates tabs (Stacks / Cross-folder) restyled to pill row in the
  Toolbar to match /tags; tab state moved into the route and bound to
  ?tab=...
- New aggregateKeywords() service fans out per-photo getPhoto calls so
  user-typed Details.Keywords surface on /tags (PhotoPrism's /labels
  only returns classifier output).
- RightSidebar renders photo.Labels[] as dashed-border chips after the
  Keywords section, each linking to /?q=label:slug.
- /colors and /ratings routes redirect to /tags?tab=colors|ratings so
  old bookmarks still land somewhere useful; LeftSidebar drops their
  entries and the Tags badge now sums labels + ratings + colors.
- listFolderCounts dedupes by UID (merged=false returns one row per
  FILE, so HEIC+JPG / Live Photo / RAW+JPG pairs were inflating folder
  counts ~2x).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

web/src/lib/services/photoprism.ts

@@ -2,6 +2,7 @@ import axios, { AxiosError, type AxiosInstance } from 'axios';
 import { browser } from '$app/environment';
 import { goto } from '$app/navigation';
 import { adoptSession, clearSession, session } from '$lib/stores/session.svelte';
+import { primaryFile } from '$lib/types/photoprism';
 import type {
 	PpClientConfig,
 	PpPhoto,
@@ -385,6 +386,62 @@ export interface PpLabel {
 	Thumb?: string;
 }
 
+/**
+ * User-set keyword aggregation. PhotoPrism's `/labels` endpoint only
+ * surfaces classifier-derived labels; user-typed keywords live in
+ * `Details.Keywords` (a flat comma-separated string) which is *not*
+ * included in the `/photos` list response. The only path to it is the
+ * single-photo endpoint, so we fetch the library top-end (capped at
+ * PhotoPrism's 1000-row ceiling), fan out `getPhoto` calls in batches,
+ * and aggregate.
+ *
+ * Slow but cached upstream via TanStack — keys under the `['photos', …]`
+ * prefix so the existing photo-mutation invalidations cascade to it.
+ */
+export interface AggregatedKeyword {
+	keyword: string;
+	count: number;
+	/** UID of an arbitrary photo carrying this keyword — used as the
+	 *  thumbnail source so the tile is visually consistent with
+	 *  classifier-label tiles. */
+	sampleUid: string;
+	sampleHash: string;
+}
+
+export async function aggregateKeywords(): Promise<AggregatedKeyword[]> {
+	const list = await listPhotos({ count: 1000, merged: true });
+	const buckets = new Map<string, AggregatedKeyword>();
+	const concurrency = 8;
+	for (let i = 0; i < list.length; i += concurrency) {
+		const slice = list.slice(i, i + concurrency);
+		const fulls = await Promise.all(
+			slice.map((p) => getPhoto(p.UID).catch(() => null))
+		);
+		for (let j = 0; j < slice.length; j++) {
+			const photo = slice[j];
+			const full = fulls[j];
+			if (!full) continue;
+			const raw = full.Details?.Keywords ?? '';
+			if (!raw) continue;
+			for (const kw of raw.split(',').map((k) => k.trim()).filter(Boolean)) {
+				const bucket = buckets.get(kw);
+				if (bucket) {
+					bucket.count++;
+					continue;
+				}
+				const hash = photo.Hash ?? primaryFile(photo).Hash;
+				buckets.set(kw, {
+					keyword: kw,
+					count: 1,
+					sampleUid: photo.UID,
+					sampleHash: hash
+				});
+			}
+		}
+	}
+	return Array.from(buckets.values()).sort((a, b) => b.count - a.count);
+}
+
 export async function listLabels(): Promise<PpLabel[]> {
 	// `all=true` includes labels PhotoPrism has soft-deleted (auto-hidden
 	// low-confidence classifier hits, manually-removed labels). They're