capture: close dialog on Archive; rich per-URL and batch toast notifications (#22)

* http: realistic UA; fall back to WebPage on probe failure

Replace bare 'archivr/0.1' user agent with a full Chrome 131 UA string
(archivr/0.1 token retained at the end) in both probe_url_kind and download.

More importantly, stop hard-failing when probe_url_kind returns an error.
Sites behind Cloudflare's managed JS challenge (e.g. Medium) return 403
before any header tuning can help — a plain HTTP client cannot pass the
challenge. Instead, log a warning and fall back to Source::WebPage so that
the SingleFile/Chromium path gets a chance; a real browser can solve the
challenge transparently.

* capture: close dialog on Archive; rich per-URL and batch toast notifications

UX changes:
- Pressing Archive closes the capture dialog immediately; jobs continue
  polling in the background (component stays mounted).
- Probe failures (e.g. Cloudflare JS challenge, HTTP 403) now fall back to
  Source::WebPage so SingleFile/Chromium can attempt the capture instead of
  hard-failing at the probe stage.

Toast notifications:
- Single URL: green 'Archived' on success, amber 'Archived with warnings'
  (with expandable detail) when uBlock/cookie-ext was skipped, red 'Capture
  failed' with expandable error text on failure. Per-item warning/error toasts
  include the locator so the user knows which URL was affected.
- Multi-URL batch: per-item failure and warning toasts still fire with
  locators; per-item success toasts are suppressed. Once all jobs settle a
  single summary toast fires: 'N archived', 'N archived (M with warnings)',
  'N archived (M with warnings), F failed', or 'N failed'. The summary Detail
  section lists the exact URLs that failed or warned, so the user retains that
  information after per-item toasts auto-dismiss.
- Batch summary color: green = all clean; amber = any warnings or failures
  present; red = all failed.
- handleIgnoreUblock now only removes per-item warning toasts (those with a
  locator) and persists the ignore flag; batch summary warnings (locator=null)
  are not swept.

ToastStack improvements:
- All three branches (success/warning/error) support a toast.headline field
  so batch summaries can set their own copy.
- Warning branch: Details button conditional on toast.text; Ignore button
  conditional on toast.locator (uBlock-specific, not shown on batch summaries).
- Locator display uses hostname/…/last-segment for URLs (preserves domain for
  context, tail for identity) and tail-truncation for shorthands; full locator
  in title attribute for hover.
- Icon colors: green for success (✓), amber for warning (⚠), red for error (✕).

crates/archivr-core/src/capture.rs

@@ -900,14 +900,12 @@ pub fn perform_capture(
             Ok(downloader::http::UrlKind::Html) => source = Source::WebPage,
             Ok(downloader::http::UrlKind::File) => {}
             Err(e) => {
-                // Record a failed item using the pre-probe source_kind so
-                // failed_count increments and the item carries error_text.
-                let (probe_sk, probe_ek, _) = source_metadata(Source::Url);
-                let item = database::create_archive_run_item(
-                    &conn, run.id, None, 0, locator, None, probe_sk, probe_ek,
-                )?;
-                let msg = format!("Failed to probe URL: {e}");
-                return Err(fail_run(&conn, &run, &item, &msg));
+                // Probe failed (e.g. Cloudflare JS challenge, 403, network
+                // issue). Fall back to treating the URL as an HTML page so
+                // that the SingleFile/Chromium path can try — a real browser
+                // can pass bot challenges that a plain HTTP client cannot.
+                eprintln!("warn: probe failed for {locator}, assuming HTML: {e}");
+                source = Source::WebPage;
             }
         }
     }

crates/archivr-core/src/downloader/http.rs

@@ -20,7 +20,7 @@ pub enum UrlKind {
 pub fn probe_url_kind(url: &str, cookies: &HashMap<String, String>) -> Result<UrlKind> {
     let client = reqwest::blocking::Client::builder()
         .redirect(reqwest::redirect::Policy::limited(10))
-        .user_agent("archivr/0.1")
+        .user_agent("Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36 archivr/0.1")
         .build()
         .context("failed to build HTTP client")?;
 
@@ -86,7 +86,7 @@ pub fn probe_url_kind(url: &str, cookies: &HashMap<String, String>) -> Result<Ur
 pub fn download(url: &str, store_path: &Path, timestamp: &str, cookies: &HashMap<String, String>) -> Result<(String, String, Option<String>)> {
     let client = reqwest::blocking::Client::builder()
         .redirect(reqwest::redirect::Policy::limited(10))
-        .user_agent("archivr/0.1")
+        .user_agent("Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36 archivr/0.1")
         .build()
         .context("failed to build HTTP client")?;