fill_prof_email: --hardware + --platform also flow into the hydration body

The hydration template hardcoded "Apple M5 Max (40-core GPU),
OpenMM Metal backend" and "friend's M5 Max overnight" — the
binding template already honored --hardware/--platform but the
hydration one ignored them. Today's pilot-3 run on Henry's
MacBook CPU produced the right verdict (MAE 1.42 PASS) but the
generated email body was wrong about where it ran.

Fix: thread hardware_hint + platform_hint into HYDRATION_TEMPLATE
the same way they go into BINDING_TEMPLATE; soften the intro
sentence to drop the M5-Max-specific phrasing.

10/10 regression covers the new "hardware lands in body" test +
verifies the M5 Max hardcoding can never silently regress
(asserts "Apple M5 Max" is NOT in the override output).

Now I can re-generate Henry's prof email with the correct hardware
("Henry's MacBook Pro M-series, CPU platform") instead of hand-
editing the draft.

Author: Rockman6

scripts/fill_prof_email.py

@@ -42,12 +42,12 @@
 Hi {prof_name},
 
 Quick update on Milestone A. The CellSim Layer-1.3 alchemical FEP
-pipeline is now end-to-end on a Mac and the FreeSolv-12 hydration
-gate ran on a friend's M5 Max overnight. Results below.
+pipeline is now end-to-end and the FreeSolv-12 hydration gate has
+run. Results below.
 
 == What ran ==
 
-- Hardware: Apple M5 Max (40-core GPU), OpenMM Metal backend
+- Hardware: {hardware_hint} (OpenMM {platform_hint} backend)
 - Force field: OpenFF Sage 2.1.0 (small molecules) + AM1-BCC partial
   charges + TIP3P water — no learned surrogate at any layer
 - MD: openmmtools.alchemy + GHMC integrator (Metropolised Langevin),
@@ -399,6 +399,8 @@ def main(argv: list[str] | None = None) -> int:
         acetamide_sign = _grab_compound_sign(md, "acetamide") or "<missing>"
         filled = HYDRATION_TEMPLATE.format(
             prof_name=args.prof_name,
+            hardware_hint=args.hardware,
+            platform_hint=args.platform,
             overall_verdict=overall,
             n_ok=n_ok,
             n_total=n_total,

tests/fep/test_fill_prof_email_smoke.py

@@ -223,6 +223,38 @@ def test_prof_name_override():
     assert "Hi [Prof]," not in out
 
 
+def test_hydration_hardware_override_lands_in_body():
+    """--hardware / --platform must reach the hydration body, not
+    just the binding one. Earlier the hydration template had M5 Max
+    hardcoded; pilot-3 CPU run on Henry's MacBook exposed the
+    hardcoding and pushed a parametrisation."""
+    with tempfile.TemporaryDirectory(
+            prefix="cellsim_fill_") as tmp:
+        tmp = Path(tmp)
+        _write_report_with_overrides(
+            tmp, kind="hydration", verdict="PASS",
+            rows_table=(
+                "| methane | `C` | +2.00 | +1.78 | 0.15 | -0.22 | "
+                "0.22 | ✓ | 4290 | |\n"
+                "| acetamide | `CC(=O)N` | -9.71 | -8.90 | 0.55 | "
+                "+0.81 | 0.81 | | 560 | |"))
+        rc, out = _run_fill(
+            tmp,
+            "--hardware", "Henry's MacBook Pro M-series, 16 GB RAM",
+            "--platform", "CPU (Reference/CPU only; no Metal)")
+    assert rc == 0, out
+    assert "Henry's MacBook Pro M-series, 16 GB RAM" in out, (
+        "hydration body must use --hardware override; got:\n"
+        f"{out[:800]}")
+    assert "OpenMM CPU (Reference/CPU only; no Metal) backend" in out, (
+        "hydration body must use --platform override; got:\n"
+        f"{out[:800]}")
+    # The old M5 Max hardcoding must be gone for the hydration path.
+    assert "Apple M5 Max" not in out, (
+        "hydration template still has M5 Max hardcoded; got:\n"
+        f"{out[:800]}")
+
+
 def test_fail_case_fixture_hydration():
     """End-to-end on the committed fail_case fixture — methane
     sign flipped, MAE 1.67. Email reflects both failures."""
@@ -262,6 +294,7 @@ def test_fail_case_fixture_hydration():
         test_missing_numbers_returns_exit_2,
         test_no_report_md_returns_exit_1,
         test_prof_name_override,
+        test_hydration_hardware_override_lands_in_body,
         test_fail_case_fixture_hydration,
     ]
     fails = []