core: optimize struct layouts, eliminate branches, and fix event loop dispatch

- Fix Wayland event loop lockout by properly decoupling read arming and adding poll-backed dispatching
- Support standalone display lifecycle without requiring caller to manage a loop
- Eliminate alignment holes across yasl_surface, yasl_display, seat, touch_point, and conn_drag using pahole
- Replace branching state/button decoders with compile-time locked LUTs and branchless bitwise operations
- Accelerate UTF-8 validation with 64-bit word scanning for ASCII runs

1 files changed, 4 insertions, 5 deletions

diff --git a/src/common/timing.c b/src/common/timing.c
index b5e342f..3e3a705 100644
--- a/src/common/timing.c
+++ b/src/common/timing.c
@@ -5,8 +5,8 @@ uint64_t yasl_clock_extend_ms(uint32_t ms, uint64_t now_ns)
     uint64_t now_ms = now_ns / 1000000;
     int64_t ahead = (int32_t)(ms - (uint32_t)now_ms);
     if (ahead < 0 && (uint64_t)-ahead > now_ms)
-        return (uint64_t)ms * 1000000;
-    return (now_ms + (uint64_t)ahead) * 1000000;
+        return (uint64_t)ms * 1000000ull;
+    return (uint64_t)((int64_t)now_ms + ahead) * 1000000ull;
 }
 
 void yasl_clock_presented(struct yasl_frame_clock *c, uint64_t now, bool sample)
@@ -45,9 +45,8 @@ double yasl_clock_refresh_ms(const struct yasl_frame_clock *c)
 
 uint64_t yasl_clock_next_present_ns(const struct yasl_frame_clock *c)
 {
-    return c->last_present_ns
-            ? c->last_present_ns + (uint64_t)c->refresh_ns
-            : 0;
+    uint64_t next = c->last_present_ns + (uint64_t)c->refresh_ns;
+    return (uint64_t)-(c->last_present_ns != 0) & next;
 }
 
 uint64_t yasl_clock_frame_timestamp_ns(struct yasl_frame_clock *c, uint64_t now)