std.sync.atomic: extended atomic helper functions (#8866)

- deprecates `std.Thread.spinLoopHint` and moves it to `std.atomic.spinLoopHint`
- added an Atomic(T) generic wrapper type which replaces atomic.Bool and atomic.Int
- in Atomic(T), selectively expose member functions depending on T and include bitwise atomic methods when T is an Integer
- added fence() and compilerFence() to std.atomic

lib/std/Thread/Condition.zig

@@ -115,7 +115,7 @@ pub const AtomicCondition = struct {
                             else => unreachable,
                         }
                     },
-                    else => spinLoopHint(),
+                    else => std.atomic.spinLoopHint(),
                 }
             }
         }

lib/std/Thread/Mutex.zig

@@ -126,7 +126,7 @@ pub const AtomicMutex = struct {
 
             var iter = std.math.min(32, spin + 1);
             while (iter > 0) : (iter -= 1)
-                std.Thread.spinLoopHint();
+                std.atomic.spinLoopHint();
         }
 
         new_state = .waiting;
@@ -149,7 +149,7 @@ pub const AtomicMutex = struct {
                         else => unreachable,
                     }
                 },
-                else => std.Thread.spinLoopHint(),
+                else => std.atomic.spinLoopHint(),
             }
         }
     }

lib/std/Thread/StaticResetEvent.zig

@@ -182,7 +182,7 @@ pub const AtomicEvent = struct {
                 timer = time.Timer.start() catch return error.TimedOut;
 
             while (@atomicLoad(u32, waiters, .Acquire) != WAKE) {
-                std.os.sched_yield() catch std.Thread.spinLoopHint();
+                std.os.sched_yield() catch std.atomic.spinLoopHint();
                 if (timeout) |timeout_ns| {
                     if (timer.read() >= timeout_ns)
                         return error.TimedOut;
@@ -293,7 +293,7 @@ pub const AtomicEvent = struct {
                         return @intToPtr(?windows.HANDLE, handle);
                     },
                     LOADING => {
-                        std.os.sched_yield() catch std.Thread.spinLoopHint();
+                        std.os.sched_yield() catch std.atomic.spinLoopHint();
                         handle = @atomicLoad(usize, &event_handle, .Monotonic);
                     },
                     else => {