fix(core): read pod cgroup limits instead of node limits in resource metrics

Author: leosvelperez

Cargo.lock

@@ -75,7 +75,17 @@ static_assertions = "1.1"
 wrap-ansi = "0.1"
 
 [target.'cfg(windows)'.dependencies]
-winapi = { version = "0.3", features = ["fileapi", "psapi"] }
+winapi = { version = "0.3", features = [
+    "fileapi",
+    "psapi",
+    # Used by the metrics collector to detect Job Object CPU rate, affinity,
+    # and memory limits — Windows analog to the Linux cgroup detection.
+    "jobapi",
+    "jobapi2",
+    "processthreadsapi",
+    "winbase",
+    "winnt",
+] }
 
 [target.'cfg(windows)'.build-dependencies]
 winres = "0.1"
@@ -92,6 +102,11 @@ uuid = "1"
 mio = "1.0"
 nix = { version = "0.30.0", features = ["fs", "process", "signal"] }
 
+# Used by the metrics collector to read sched_getaffinity directly, bypassing
+# std::thread::available_parallelism's floor-rounded cgroup quota handling.
+[target.'cfg(target_os = "linux")'.dependencies]
+libc = "0.2"
+
 [target.'cfg(not(target_arch = "wasm32"))'.dependencies]
 arboard = { version = "3.4.1", features = ["wayland-data-control"] }
 crossterm = { version = "0.29.0", features = ["event-stream", "use-dev-tty"] }