fix(rules): avoid the firing-into-melee penalty at 10 feet

SRD 'Attack - Shooting or Throwing into a Melee': a target at least 10 ft from the nearest friendly character lets you avoid the -4 penalty entirely (0, not -2). The -2 is reserved for targets two size categories larger than the engaged friendlies (0 at three+), which is not modeled (documented). Golden transcript unaffected (shot still misses).

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>

tests/test_accuracy_rules.py

@@ -337,7 +337,11 @@ def test_firing_into_melee_minus4() -> None:
     assert t.hp == 100
 
 
-def test_firing_into_melee_minus2_when_target_10ft_away() -> None:
+def test_firing_into_melee_avoided_when_target_10ft_away() -> None:
+    # PF1e (SRD "Attack" — Shooting or Throwing into a Melee): "If your target
+    # [...] is at least 10 feet away from the nearest friendly character, you
+    # can avoid the -4 penalty" — the 10+ ft clause is 0, not -2 (the -2 is
+    # reserved for targets two size categories larger, not modeled).
     archer = _combatant("archer", ac=99, attacks=(_bow("archer", attack_bonus=8),))
     tgt = _combatant("tgt", ac=15, attacks=(_bow("tgt"),))
     guard = _combatant("guard", ac=99)
@@ -349,8 +353,8 @@ def test_firing_into_melee_minus2_when_target_10ft_away() -> None:
     )
     a, t, _g = eng._states
     result = eng.resolve_attack(a, t, a.combatant.attacks[0])
-    assert result.total == 15
-    assert result.melee_penalty == -2
+    assert result.total == 17
+    assert result.melee_penalty == 0
     assert result.hit is True
     assert result.damage == 3