Correct the learned-heuristic timings, and say what gripper's seed does

The README's transfer table quoted 0.038 / 0.561 / 0.169 seconds. The
cache the published Research view renders from -- promo/rl-data.json,
transfer_after -- says 0.0365 / 0.5337 / 0.1542. Expansions and costs in
the same three rows match the cache exactly, so the wall-clock column
alone came from a run nobody kept. A test pins the page to that cache;
nothing pinned this table, so it drifted.

generate_gripper takes a seed and, alone among the generators, never
builds an rng from it: in gripper every ball starts in room A, so size
and grippers determine the instance completely and the seed reaches
only the problem name. Worth a docstring rather than a fix, because the
trap is silent -- a train/test split that varies the seed at a fixed
size hands you the same instance twice.

Author: guilyx

README.md

@@ -344,13 +344,15 @@ seen, against greedy best-first search:
 
 | heuristic | coverage | expansions | seconds | plan cost |
 |---|---|---|---|---|
-| **`learned`** | 1.00 | **137** | **0.038** | **48.2** |
-| `hff` | 1.00 | 518 | 0.561 | 51.8 |
-| `goalcount` | 1.00 | 2483 | 0.169 | 51.0 |
-
-Nearly four times fewer expansions than `hff` and fifteen times faster, because
-the network is a thousand multiply-adds and `hff` is a relaxed-plan extraction.
-Those are the numbers the command above prints, on one CPU, in about a minute.
+| **`learned`** | 1.00 | **137** | **0.036** | **48.2** |
+| `hff` | 1.00 | 518 | 0.534 | 51.8 |
+| `goalcount` | 1.00 | 2483 | 0.154 | 51.0 |
+
+Nearly four times fewer expansions than `hff` and about fifteen times faster,
+because the network is a thousand multiply-adds and `hff` is a relaxed-plan
+extraction. These are read from `promo/rl-data.json`, the cache the published
+[Research view](https://openplan-labs.github.io/PythonPDDL/) renders from, so
+the page and this table cannot drift apart.
 
 It does not always win — on logistics it loses to `hff` by 6× and the reason is
 exact rather than mysterious: that domain has two predicates, so the feature

jupyddl/generator.py

@@ -164,7 +164,17 @@ def generate_blocksworld(size: int = 6, seed: int = 0):
 
 
 def generate_gripper(size: int = 4, seed: int = 0, grippers: int = 2):
-    """``size`` balls to move from room A to room B. Plan length grows linearly."""
+    """``size`` balls to move from room A to room B. Plan length grows linearly.
+
+    Unlike every other generator here, ``seed`` does not vary the instance:
+    in gripper every ball starts in room A and must end in room B, so
+    ``size`` and ``grippers`` determine the problem completely. The seed
+    reaches only the problem name, which keeps generated files distinct.
+
+    The consequence is worth stating because it is easy to miss: a
+    train/test split that varies the seed and holds the size fixed produces
+    the *same instance* twice for this domain. Vary ``size`` instead.
+    """
     balls = _objects("ball", max(1, size))
     hands = _objects("gripper", max(1, grippers))
     init = ["(at-robby rooma)"]