test(e2e): scope EAI-7052 lemonade-inference xfail to any OS, not just Linux

The lemonade inference-failure xfail conditions were scoped os="linux", but the
Strix Halo Windows run shows the same failure there — lemonade reports the
service ready, then inference drops the connection ("CURL error: Failure when
receiving data from the peer" / "error sending request"). So the four broadened
inference scenarios (serve-lemonade-inference, serve-readiness-contract,
chat-tool-definitions-accepted, chat-end-to-end-local-model) read as unexpected
failures on Strix Windows. Drop the os constraint — EAI-7052 is engine-specific
(lemonade llama.cpp backend), not OS-specific — so it xfails on any lemonade host.

Signed-off-by: fredespi <fredrik.espinoza@gmail.com>

Author: fredespi

tests/e2e-cucumber/expectations.toml

@@ -62,20 +62,20 @@ bug = "EAI-7333"
 reason = "CLI reports the vLLM service ready, but an immediate inference request fails."
 serve_timeout_secs = 90
 
-# On a lemonade-default Linux host this scenario does real lemonade inference,
-# which hits the EAI-7052 Vulkan hang (see serve-lemonade-inference).
+# On a lemonade host this scenario does real lemonade inference, which hits
+# the EAI-7052 backend failure (see serve-lemonade-inference).
 [["serve-readiness-contract"]]
-when = { effective_engine = "lemonade", os = "linux" }
+when = { effective_engine = "lemonade" }
 bug = "EAI-7052"
-reason = "Lemonade uses its Vulkan llama.cpp backend instead of system ROCm; inference hangs."
+reason = "Lemonade inference fails on the llama.cpp backend (Vulkan hang on Linux; connection drop on Windows) instead of using system ROCm."
 serve_timeout_secs = 90
 
-# --- EAI-7052: lemonade falls back to Vulkan llama.cpp instead of system ROCm ---
-# and inference hangs. Observed on Linux GPU hosts serving GGUF via lemonade.
+# --- EAI-7052: lemonade llama.cpp backend fails inference instead of using ---
+# system ROCm. Both Linux (Vulkan hang) and Windows (connection drop) — any OS.
 [["serve-lemonade-inference"]]
-when = { effective_engine = "lemonade", os = "linux" }
+when = { effective_engine = "lemonade" }
 bug = "EAI-7052"
-reason = "Lemonade uses its Vulkan llama.cpp backend instead of system ROCm; inference hangs."
+reason = "Lemonade inference fails on the llama.cpp backend (Vulkan hang on Linux; connection drop on Windows) instead of using system ROCm."
 serve_timeout_secs = 90
 
 # --- EAI-7223: chat requests with tool definitions rejected (vLLM path) ---
@@ -85,11 +85,11 @@ bug = "EAI-7223"
 reason = "Chat requests carrying tool definitions are not accepted by the served vLLM model."
 serve_timeout_secs = 90
 
-# On a lemonade-default Linux host this does real lemonade inference → EAI-7052 hang.
+# On a lemonade host this does real lemonade inference → EAI-7052 backend failure.
 [["chat-tool-definitions-accepted"]]
-when = { effective_engine = "lemonade", os = "linux" }
+when = { effective_engine = "lemonade" }
 bug = "EAI-7052"
-reason = "Lemonade uses its Vulkan llama.cpp backend instead of system ROCm; inference hangs."
+reason = "Lemonade inference fails on the llama.cpp backend (Vulkan hang on Linux; connection drop on Windows) instead of using system ROCm."
 serve_timeout_secs = 90
 
 # --- EAI-7221: end-to-end chat through a locally served model (vLLM path) ---
@@ -99,11 +99,11 @@ bug = "EAI-7221"
 reason = "End-to-end chat through a locally served vLLM model does not return a valid reply."
 serve_timeout_secs = 90
 
-# On a lemonade-default Linux host this does real lemonade inference → EAI-7052 hang.
+# On a lemonade host this does real lemonade inference → EAI-7052 backend failure.
 [["chat-end-to-end-local-model"]]
-when = { effective_engine = "lemonade", os = "linux" }
+when = { effective_engine = "lemonade" }
 bug = "EAI-7052"
-reason = "Lemonade uses its Vulkan llama.cpp backend instead of system ROCm; inference hangs."
+reason = "Lemonade inference fails on the llama.cpp backend (Vulkan hang on Linux; connection drop on Windows) instead of using system ROCm."
 serve_timeout_secs = 90
 
 # --- EAI-7383: `rocm help` lists subcommands in declaration order, not