close guardrail gaps and fix the scoring pipeline top to bottom

Constraint resolver: aggregate mass/footprint across a combo instead of
pairwise-only checks, treat medium/atmosphere as agreement not supply/demand,
reduce multi-provider checks by best/sum instead of AND-ing every provider,
fail closed on unrecognized mutex values, add a propulsion-viability
(thrust-to-weight) rule. Seed data updated to match (nuclear/solar-sail
footprint floors, water-medium exclusions, explicit ground/gravity providers).

Domain metric units were stored globally per metric name instead of
per-domain, silently corrupting cost_efficiency for every domain but the
first one seeded — fixed with a schema migration.

Stub estimator's cost_efficiency/safety/availability/reliability were a
backwards formula and flat constants; replaced with heuristics grounded in
each entity's thrust_profile/energy_form/infrastructure.

LLM estimate_physics() now receives each metric's unit and expected range
instead of a bare name, fixing wildly miscalibrated estimates traced back to
the prompt's own hardcoded example anchoring the model to the wrong order of
magnitude. Sharpened the safety-estimation and plausibility-review prompts.
Deduped provider parsing logic into llm/parsing.py.

Web pipeline form can now pick an LLM provider per run instead of only via
server env var.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

src/physcom/llm/registry.py

@@ -7,32 +7,40 @@ import os
 from physcom.llm.base import LLMProvider
 
 
-def build_llm_provider() -> LLMProvider | None:
-    """Return an LLMProvider based on env vars, or None if not configured.
+def build_llm_provider(
+    provider: str | None = None,
+    model: str | None = None,
+    host: str | None = None,
+) -> LLMProvider | None:
+    """Return an LLMProvider, or None if not configured.
+
+    Explicit args (e.g. from a per-request web form) override env vars;
+    passing nothing falls back to the env-var-only behavior below.
 
     LLM_PROVIDER   — provider name ('gemini', 'ollama'; more can be added)
-    GEMINI_API_KEY — required when LLM_PROVIDER=gemini
+    GEMINI_API_KEY — required when provider is 'gemini' (server env only,
+                     never accepted as a request param)
     GEMINI_MODEL   — optional Gemini model name (default: gemini-2.0-flash)
     OLLAMA_MODEL   — optional Ollama model name (default: qwen2.5:7b)
     OLLAMA_HOST    — optional Ollama server URL (default: http://localhost:11434)
     """
-    provider = os.environ.get("LLM_PROVIDER", "").lower().strip()
+    provider = (provider or os.environ.get("LLM_PROVIDER", "")).lower().strip()
 
-    if not provider:
+    if not provider or provider == "stub":
         return None
 
     if provider == "gemini":
         api_key = os.environ.get("GEMINI_API_KEY", "")
         if not api_key:
-            raise ValueError("LLM_PROVIDER=gemini requires GEMINI_API_KEY to be set")
-        model = os.environ.get("GEMINI_MODEL", "gemini-2.0-flash")
+            raise ValueError("Gemini requires GEMINI_API_KEY to be set in the server environment")
+        model = model or os.environ.get("GEMINI_MODEL", "gemini-2.0-flash")
         from physcom.llm.providers.gemini import GeminiLLMProvider
         return GeminiLLMProvider(api_key=api_key, model=model)
 
     if provider == "ollama":
-        model = os.environ.get("OLLAMA_MODEL", "qwen2.5:7b")
-        host = os.environ.get("OLLAMA_HOST", "http://localhost:11434")
+        model = model or os.environ.get("OLLAMA_MODEL", "qwen2.5:7b")
+        host = host or os.environ.get("OLLAMA_HOST", "http://localhost:11434")
         from physcom.llm.providers.ollama import OllamaLLMProvider
         return OllamaLLMProvider(model=model, host=host)
 
-    raise ValueError(f"Unknown LLM_PROVIDER: {provider!r}. Supported: gemini, ollama")
+    raise ValueError(f"Unknown LLM provider: {provider!r}. Supported: gemini, ollama, stub")