udev/net: implement IRQAffinityPolicy=spread with topology awareness

Implement the spread policy for IRQ affinity distribution using a
topology-aware algorithm. The algorithm:

1. Discovers CPU topology from sysfs (NUMA node, package, die/L3, core)
2. Groups CPUs by L3 cache domain (die) with equidistant ordering
3. Round-robins across dies, spreading IRQs across the system
4. Uses first hyperthread of each core before second hyperthreads
5. Applies IRQ affinity via /proc/irq/<n>/smp_affinity

When there are more IRQs than CPUs, queues wrap around using round-robin.

src/basic/sort-util.c

@@ -81,3 +81,12 @@ int cmp_uint16(const uint16_t *a, const uint16_t *b) {
 
         return CMP(*a, *b);
 }
+
+int cmp_unsigned(const unsigned *a, const unsigned *b) {
+        /* This is called from qsort()s inner loops. Correctly implemented qsort will never pass NULL so we
+           just suppress the check via POINTER_MAY_BE_NULL instead of assert() to avoid the runtime cost. */
+        POINTER_MAY_BE_NULL(a);
+        POINTER_MAY_BE_NULL(b);
+
+        return CMP(*a, *b);
+}

src/basic/sort-util.h

@@ -44,3 +44,4 @@ void qsort_r_safe(void *base, size_t nmemb, size_t size, comparison_userdata_fn_
 
 int cmp_int(const int *a, const int *b);
 int cmp_uint16(const uint16_t *a, const uint16_t *b);
+int cmp_unsigned(const unsigned *a, const unsigned *b);