fix(bypass): keep Anna's Archive's aa_ddg_check so clearance replays (#1305)

## What

Add `aa_ddg_check` to the cookie-store allowlist. One name, one test.

## Why

Every replay of stored clearance ends in the `?check=1` redirect loop,
so each search pays a fresh browser solve. On this instance (v1.3.15,
WireGuard egress, 0 VPN restarts across the traces) not one replay was
accepted in three days of DEBUG logs.

The `__ddg*` cookies are stored and replayed correctly. Anna's Archive
also sets a cookie of its own, `aa_ddg_check`, and its `?check=1` hop
only answers with the page when that cookie is present too. The
allowlist keeps `cf_*` and `__ddg*` names, so this one was never stored.

## Measured, same egress IP, cookies taken from one solve

| replayed | plain `requests` | `curl_cffi`, Chrome TLS fingerprint |
|---|---|---|
| filtered `__ddg*` only (current behaviour) | 302 → 302 → 302 … loop |
302 → 302 → 302 … loop |
| filtered + `__ddg8_/9_/10_` | loop | loop |
| filtered + `aa_ddg_check` | **302 → 200, real search page** | 302 →
200 |
| `aa_ddg_check` alone | 302 → 403 | — |

So the TLS fingerprint is not the problem, the per-check trio is not the
answer, and the cookie needs the `__ddg*` clearance next to it. Cookie
attributes as issued: domain `.annas-archive.gl`, path `/`, expiry 90
days. It is not bound to the query, and it is accepted with a stock
Python User-Agent.

## Through the real fetch path

Same process, `html_get_page`, the name allowlisted, three different
queries:

```
1st: solve expected         25.8s  bypass_calls=1  title='frankenstein shelley - search - an'  md5=True
2nd: other query             9.5s  bypass_calls=0  title='pride and prejudice austen - searc'  md5=True
3rd: third query             4.8s  bypass_calls=0  title='dracula stoker - search - anna's a'  md5=True
```

## Notes

- `tests/bypass/test_ddg_cookie_reuse.py` gains
`test_aa_check_cookie_is_stored`; its docstring table gains the row. The
bypass tests need seleniumbase to import and do not run on my macOS
host, so this leans on CI. `ruff check` and `ruff format --check` pass.
The logic was checked directly against `cookie_store` with the settings
registry stubbed.
- `__ddgmark_` carries a 24 h expiry, so the store's clearance is good
for about a day before the next solve, which is what a browser would see
too.
- Follow-up to #1286. Same instance, same method: DEBUG trace, then a
probe script inside the container.

shelfmark/bypass/cookie_store.py

@@ -38,6 +38,10 @@ DDG_COOKIE_NAMES = {
     "ddg_last_challenge",
 }
 
+# Anna's Archive's own pass for its ?check=1 hop. Without it the hop 302s back
+# forever, however good the __ddg* clearance is.
+AA_COOKIE_NAMES = {"aa_ddg_check"}
+
 # DDoS-Guard cookies that describe *one* check rather than granting clearance, and so
 # must never be replayed on a later request. Observed live on Anna's Archive:
 #
@@ -88,7 +92,8 @@ def _should_extract_cookie(name: str, *, extract_all: bool) -> bool:
         return True
     is_cf = name in CF_COOKIE_NAMES or name.startswith("cf_")
     is_ddg = name in DDG_COOKIE_NAMES or name.startswith("__ddg")
-    return is_cf or is_ddg
+    is_aa = name in AA_COOKIE_NAMES
+    return is_cf or is_ddg or is_aa
 
 
 def _cookie_field(cookie: Any, name: str) -> Any:

tests/bypass/test_ddg_cookie_reuse.py

@@ -1,11 +1,12 @@
 """DDoS-Guard cookie reuse between requests.
 
-Anna's Archive issues nine cookies after a solve, and they are not equivalent:
+Anna's Archive issues ten cookies after a solve, and they are not equivalent:
 
     __ddg1_/__ddg2_/__ddgid_   ~1 year   clearance
     __ddgmark_                 ~1 day
     __ddg5_                    session
     __ddg8_/__ddg9_/__ddg10_   ~40 min   one check: token, CLIENT IP, TIMESTAMP
+    aa_ddg_check               ~90 days  Anna's Archive's own pass for its ?check=1 hop
 
 Replaying the last three is what produces the ?check=1 redirect loop. They describe a
 single check, so once the timestamp ages out - or the egress IP changes, routine
@@ -97,6 +98,15 @@ def test_per_check_cookies_are_not_stored():
         assert ephemeral not in stored
 
 
+def test_aa_check_cookie_is_stored():
+    """Without aa_ddg_check the ?check=1 hop loops, whatever __ddg* is replayed."""
+    _store([_Cookie("__ddg1_", "a"), _Cookie("__ddg5_", "b"), _Cookie("aa_ddg_check", "ok")])
+
+    stored = ib.get_cf_cookies_for_domain("annas-archive.gl")
+
+    assert stored == {"__ddg1_": "a", "__ddg5_": "b", "aa_ddg_check": "ok"}
+
+
 def test_clearance_cookies_survive():
     _store([_Cookie("__ddg1_", "a"), _Cookie("__ddg2_", "b"), _Cookie("__ddgid_", "c")])